During the selection of equipment on the Collahuasi project in 2001, Kvaerner’s lead engineer for the grinding area kept repeating that Cadia had “dodged a bullet”.  This hyperbolical phrase reflected the fear that had been generated in the mineral processing industry due to the events at Cadia.  The mine owners’ and engineering contractors’ reactions on Collahuasi and several other projects were extreme, as many would not consider a Siemens motor even though the Cadia motor had been repaired and had been operating for more than 3 years.  In most cases, the die had been cast prior to any tendering and ABB cornered the market.  Unfortunately for Collahuasi (and others), their desire to avoid “the bullet” despite the fact that the Cadia issue had been resolved, resulted in a much greater problems.  What is worse though is that we can definitely say with hindsight that there was no bullet to dodge at Cadia.

The original Cadia stator design was only just too flexible. It wasn’t “floppy” and impossible to manage or control. The maximum amplitude of the vibrations during a vibration excursion was about 7 mm.  This was a horizontal deflection at the top of the frame. The greatest reduction of the air-gap was about 5 mm at ±45º to the horizontal.  Also, the stator was metastable, not unstable. It switched from one stable state (normal operation) to another stable state (albeit with higher vibration levels) and back again.  When the motor vibrated, the vibrations were self-limiting due to the magnetic forces in the air-gap, unlike resonant vibration where only the material damping limits the vibration level.  There was physical evidence that the vibrations were metastable. The stator never touched the rotor during the early days of operation despite several extended periods of vibration excursions that had occurred, usually on cold nights.  The vibrations did not increase and decrease proportionally with speed.  The onset of vibration was sudden at a particular speed range unlike the gradual increase that would be expected if the motor was resonating as a linear system.  Once the vibrations had started, the motor had to be stopped or its speed had to be reduced dramatically before they would cease.  Again, this was not a resonant behaviour.  We did not completely understand the metastable nature of the vibrations at the time. This understanding was developed once we solved the mathematical equations that governed the stator motion.  Whilst we had the mathematical model of the vibrations in 1998, we were only able to solve it approximately because the equations were highly nonlinear.  More refined solutions that enabled assessments of new designs were developed after the Strongback* was installed at Cadia.  The solutions to the mathematical models show that the vibrations quickly increase from a very low level to about 300 mm/s and remain at that level until the speed of the motor is raised or lowered outside the critical speed range around 9 rpm. The vibrations then quickly reduce to the normal operating levels for both lower and higher speeds.

There is no doubt that the situation at Cadia was worrisome for Newcrest and Siemens.  It was quite scary to see the motor suddenly change its behaviour.  The forces generated by the vibration were sufficient to shake the entire grinding plant.  However, quite soon after the behaviour was detected, Siemens worked out that the motor could be wedged to the foundations to prevent the vibrations from occurring at all.  This was implemented by the plant’s engineering team and it worked very well as a temporary measure to prevent vibrations and hence any significant loss of production until it was agreed that the stiffener should be installed.**

So, there was no bullet to dodge at Cadia, as the stator was never a danger of collapsing onto the rotor.  A more appropriate metaphor would have been that a gun had been waved around when the motor was commissioned and this was frightening because no-one knew that it wasn’t loaded.  The danger of a catastrophic motor failure was perceived rather than real.

A Design Tool for Motors

Whilst the most exciting part of the work at Cadia was developing the repair, the most beneficial technical output of the work to our business was developing a design tool to assess new motor installations.  It was evident early on that the nonlinear forces in the air-gap resulted in a sub-harmonic vibration.  Sub-harmonic excitation could be modelled using Hill or Mathieu equations but knowing this did not help much in assessing new machines, i.e., the knowledge that sub-harmonic vibrations could occur does not help in predicting if they will occur.  So,we initially came up with the 3 Hz design rule that related the operating natural frequency of the motor with the maximum pole passing frequency, i.e., the operating natural frequency had to be 3 Hz greater than half the pole-passing frequency.  This was a robust tool but it resulted in a conservative assessment.  Later, we developed an accurate mathematical model that simulated the stator behaviour at Cadia without any arbitrary adjustment of the input variables.  This was used successfully on many large Siemens drives including the largest drive at Sino Iron (40 ft, 28 MW).  The design tool requires input from the vendor in the form of the magnetic pull curve for each pole.  The accuracy of these curves defines the accuracy of the model and it was evident from experiments at Cadia that the magnetic pull curve developed during design a motor can be substantially different from the actual curves and so a safety factor has to be included in the calculation. Despite this, the design tool has produced reasonable assessments of new drives and is considered sufficiently accurate that Siemens asked that it be used to assess the Sion Iron design as an independent design check of their modelling.  The development of the design tool is the focus of this special interest topic and will be described in the next post.

NEXT POST:  Description of Problem and Modelling

*  – The stiffener EAnD developed and Siemens implemented to resolve the vibration issue.

** – Siemens tried several other options before accepting the Strongback solution

The Cadia SAG mill diameter is 40 ft and the ring motor that drives the mill is rated at 20 MW.  In 1998 when the mill and motor were commissioned both were the largest machines of their type in the world.

The stator exhibited vibration excursions that appeared to be related to the mill speed. We determined the mode shape of the vibration using approximately 25 accelerometers around the circumference of the stator and found it was a pure, in-plane motion as represented in Figure 1. The frequency of the vibration was approximately 7 Hz and it occurred then the motor reached a speed at which the pole-passing frequency was 14 Hz. The amplitude of the vibration was approximately 300 mm/s at the top of the stator horizontally.

The source of the 7 Hz vibration was not evident as the finite element model of the stator developed by Siemens during the design of the motor indicated that the natural frequency of the stator in the first in-plane mode should have been approximately 12 Hz. We measured the modal frequencies of the stator using an impact hammer when the motor was de-energised and found that the frequency of the first in-plane mode was approximately 9.5 Hz.  Whilst this result did not explain the 7 Hz vibration, it did show that the material properties that Siemens used in their analysis of the stator were incorrect.  In particular, the effective stiffness of the core was much less than had been assumed.  We then developed a separate finite element (FE) model of the stator that included a lower stiffness for the core and tuned it to the measured frequencies (Figure 1). This FE model was used to assess possible causes of the lower than expected vibration frequency and was also used to assess potential modifications to repair the stator.

The modelling and the repair of the stator are described in [1].  A stiffener or “Strongback” was added to the stator to raise its natural frequency from 9.5 to 10.7 Hz. The vibrations were eliminated. The Strongback was sized to provide a separation between the operating natural frequency and half the pole-passing frequency of at least 3 Hz. This separation equals the increase in natural frequency produced by the temporary wedges (or chocks) that were installed under the stator to control the vibration plus a safety factor (see Preface and [1]). The Strongback reduced the static deflections in the stator by more than 80%. Two pins were also added to each of the stator supports to prevent any lateral movement. Since then, the stator has performed very well.

Origin of the 7 Hz Vibration

Whilst the solution to the vibration was relatively straight forward once the behaviour or mode shape was understood, the origin of the 7 Hz vibration was not determined from the measurements or during the development of the Strongback. The FE model of the stator that was tuned to the vibration and modal measurements taken on site showed that the natural frequency of the stator would be greater than 8 Hz even if the stator core stiffness was reduced to zero.  Furthermore, it was observed during site tests that the vibration amplitude did not vary with speed as it would be expected to if the vibration were a resonance phenomenon.  Also, the amplitude was bounded and independent of speed once the vibration initiated.

The equations of motion of the stator can be expressed as:

Equation 1 is general and can be used to assess both Siemens and ABB stators for nonlinear vibration.  There is no analytical solution to the equation, so it must be solved numerically.   The solutions to mathematical equations of this sort are generally presented as stability maps. In this case, the map would be a plot of the natural frequency against q and lines would be drawn on the diagram to represent the boundaries of stable and unstable (large) vibration.  However, stability diagrams are of little use in motor design as the natural frequency and q are fixed.  It is sufficient to use direct numerical solvers to determine the solution, i.e., the vibration amplitude and frequency.

The numerical solution of equation 1 for the Cadia stator is shown in Figure 2.  When the operating frequency ω is set to the pole passing frequency of 14 Hz (equivalent to speed at which the vibrations occurred) the model predicts that the stator will initially operate normally with very low vibration amplitude but then the amplitude will begin to increase exponentially and reach a new stable state with an amplitude of approximately 7 mm. This is consistent with the measurements from site.   The large amplitude vibration will not occur if the pole-passing frequency is increased or decreased by 5% indicating that this is not a linear resonance phenomenon.  Furthermore, the amplitude of the vibration does not grow beyond 7 mm. This is because the nonlinear magnetic pull curve that initially generates the excitation of the stator also damps the vibration once the deflection reaches a certain level in a vibration cycle [2].

Assessment of New Motors

Equation 1 can be used to assess any new motor design. The magnetic pull curve defines the function f(x).  The form factors C and q can be determined from a finite element analysis of the stator, however, C requires calibration against a new machine as the magnetic pull curve provided by vendors for design are too conservative.  We determine the these form factors against measurements from the Cadia stator for Siemens motors and the Antamina stator for ABB motors .  This calibration process means that any change made to the way a new stator is modelled must also be made to the reference models of the Cadia and Antamina stators to ensure consistency.  One such change that has occurred since c.2000 when the Cadia model was first developed is the way in which Siemens calculate the magnetic pull curve.  Siemens now include electrical damping in the calculation of the magnetic pull curves for new motors and this results in lower magnetic pull forces per pole than the method they used in the original Cadia design.

The terms in equation 1 can be made more accurate by including more information from the finite element analysis of the stator in the form factors C and q but this is generally only be required in detailed design or design checks.  A simpler screening approach can be used for preliminary assessments of new designs. The numerical analysis of equation 1 yields the interesting result that sub-harmonic vibrations will only occur if the magnetic pull is sufficient to reduce the operating natural frequency of the stator to approximately half the pole-passing frequency.  This fact can be used to assess stator designs relatively simply as a first pass.  It is important to note however that impact hammer tests on the Cadia stator during operation resulted in measured natural frequencies at least 15% greater than the half-pole-passing frequency, so the simple screening approach should only be used for preliminary analysis.

Next Post:  Other Investigations of the Cadia Vibrations

References:

[1] Meimaris, C., Lai, W. K. K. L., Cox, L., 2001. Remedial design of the world’s largest SAG mill gearless drive. In Proceedings of the SAG 2001 Conference, Vancouver, Canada.

[2] Rand, R. H., Lecture notes on nonlinear vibrations. Cornell University, 2014.

We have written on risk associated with large ring motors in gearless mill drives previously [1, 2, 3].  In [1], we observed that ABB and Polysius were far too confident in 2001 when they claimed that 42 and 44 ft diameter mills and ring motors posed only manageable risks and that a 44 ft mill and GMD design was ready and waiting for a purchaser [4].  The multiple failures of ABB drives after 2001 proved this point.

A more recent contribution to the topic of risk associated with large mill drives is presented in the “Insurability of Mill Drives” [5].   The paper’s authorship is a surprising combination of an insurer’s risk engineering division and a GMD supplier (Zurich Insurance Services and ABB).  Does the inclusion of an insurer in the paper imply an endorsement of the ABB design as being an acceptable or low risk?  Does it mean that Siemens must make changes to their winding design and eliminate their tachometers so their motors become “less risky” since these issues are highlighted in the paper as being disadvantageous to ring motor design?  Let us hope not because the original Collahuasi and Antamina SAG mill stators most likely would still be operating had they been wound with the type of VPI parallel windings used by Siemens.

In this post, we review the new 24 h repair criterion proposed in [5] and how it applies to winding design.  The key items in an engineering risk assessment presented in the last section of [5] are also be reviewed.

The 24 h Repair Cycle – Potentially Dangerous

A new criterion proposed in [5] is that repairs in a GMD should only take 24 hours to complete.  These may be temporary repairs if a full repair requires an extended shut-down.  It would be dangerous for the mining industry if owners and insurers were to weight this criterion very heavily in evaluating the relative risk between motor designs.  Consider the specific example of winding design is used in the paper.  The paper indicates that the bridging of a winding can take a lot of time but this can be avoided if the winding is “designed correctly from the beginning”.  It is true that the ABB bar windings can be bridged more easily than the Siemens parallel windings.  However, the ABB windings result loads that are approximately 30% greater than those generated by Siemens windings due to the lack of electrical damping.  This is critical as the size and rated power of ring motors increase.  In [3], we showed that the Collahuasi failure occurred because the hanger plates were loaded more heavily than hanger plates in smaller motors.  Had the Collahuasi motor been wound with Siemens-type parallel windings,  the loads in the hanger plates could have been reduced by 30%.  This reduction alone would have most likely prevented the hanger plate failures at Collahuasi, as the load per hanger plate would have been less than the hanger plate loads in the Antamina and Telfer motors.  Furthermore,  ABB insulated the stator windings of the Collahuasi and Antamina motors with Nomex paper.  Had they chosen VPI insulation as used by Siemens, the frequency of partial discharges that burnt out sections of the electrical cores would have been reduced and shorting of the windings at Antamina due to lamination movement would have been avoided due the the hardness of the VPI windings.   Therefore, the Collahuasi and Antamina SAG mill stators would not have required replacement so early in their operational lives had ABB had used parallel windings with VPI insulation.  So, is a 24 h repair cycle for relatively low frequency failures of a VPI windings more important than other design features such as lower loads in the stators and the foundations?  We don’t believe so.  This “quickly repairable concept” should not rank highly in any engineering assessment of ring motors.

Burnt laminations due to winding discharge

Insurance Risk

The “Mining Industry from the Insurer’s Perspective” section in [5] is a general section on insurance issues relating to risk management and insurance rates.  This is quite informative.  However, the final section of the paper on how risk engineers evaluate GMD installations does not address the lessons learnt from the root causes of the major failures from the past 20 years.  There is nothing on:

• design capability of the vendors
• track record of specific vendors rather than GMD track records as a whole;
• the specific risk one type of ring motor design represents versus another – the motor vendors produce very different designs;
• weighting of passive protection versus active monitoring;
• short-circuit behaviour and other accident conditions;
• structural design – most failures and “near misses” have been related to structural-mechanical components;
• design culture – the root cause of not only the structural-mechanical failures but many of the electrical ones too; and
• compromises made during design to fit a plant – the motors are bespoke pieces of equipment, not off-the-shelf items.

In fact, given that the paper was written with the benefit of hind-sight, it is surprising that there is nothing in the engineering assessment list that would assist an insurer in detecting the risks associated with the major failures that have occurred in the past 20 years.

Conclusion

We have said in a previous post [3] that GMDs, ring motors and mills are designed in an evolutionary manner.  What we see as dangerous is the potential for external influences such as owners’ or insurers’ perception of risk of a fundamental aspect of a design to dramatically influence the direction of this evolution.  External influences that result in accelerated conceptual changes to GMD designs have proven to be expensive in the past [3].  Altering the winding design in a motor from parallel to series windings or vice versa would constitute a such conceptual change.  VPI insulated windings do not have a high failure frequency and therefore there is no reason to weight the time taken to bridge a winding very highly in a risk assessment.

The key items of a GMD installations listed in [5] for consideration in a risk assessment do not consider the lessons learnt from the failures of the past 20 years.

References:

[1] Meimaris, C., Price, B. F., Manchanda, S., 2006. How big is big? – Revisited. SAG 2006 Conference, Vancouver, Canada.

[2] Meimaris, C., Price, B. F., Manchanda, S., 2006. Large gearless driven mill systems – Responsibilities and capabilities. SAG 2006 Conference, Vancouver, Canada.

[3] EAnD Website Post: Literature Review: Gearless Motor Failures – A Mill Designer’s Viewpoint

[4] Riezinger, F. M., Knect, J., Patzelt, N., Errath, R. A., How big is big – Exploring today’s limits. SAG 2001 Conference, Vancouver, Canada

[5] Bos, L.,  van de Vijfeijken, M., Koponen, J., Insurability of Large Gearless Mill Drives. Link

Trunnion-mounted mills rely on castings either in grey iron for older mills, steel castings and more commonly SG iron (ductile iron) castings.  US mill vendors complain bitterly about consultants employed by the Owner for the purpose of checking the quality of these castings.  In fact, Svalbonas has written several papers on it [1, 2, 3].  Basically, his point is summarised in [1] as “[m]ining customers have never properly evaluated their structural consultants.  Being an expert theoretical analyst does not automatically qualify one for QC of castings”.  Well, better to be an expert at something we suppose.  In any case, let us look at a few examples of castings and QC.  These examples were all from the final inspection stage for owner acceptance and not part way through the finishing process.

Example 1:  Below is a video of a SAG mill casting that was offered for inspection after machining was completed.  Click on the video and note that this was typical of the castings (23 in all) supplied for two SAG mills on a specific project.

 
Figure 1: Casting offered for inspection

Example 2:  Below is a photo of one of the casting segments from the same foundry for the same mill.  Again, the casting was offered for owner final inspection after machining was completed.  The repair to the gaping hole in the flange was instigated by the owner, not the vendor.  If pinholes can reduce the fatigue strength by 40% [4], this region in its original state would have reduced the fatigue strength and hence the life of the casting by much more.

Figure 2: Mill flange defect – offered for inspection after machining completed and ready for rubber lining.

Example 3:  A casting for a mill is shown in Figure 3.  The casting was inverted from the position shown in Figure 3 when it was presented for final inspection.  The casting was in a dark corner of the factory and the flange was on blocks just high enough to enable someone to see the surface of the mating section of the flange.  When the poor surface finish was detected, the owner insisted that the casting be flipped over for a better inspection the following day.  Dark areas that were clearly dross were found in the casting flange once it was cleaned up.  The vendor’s representative said that the dark areas were not an issue as they were just “oil stains” on the metal.  It was necessary to hire an independent UT technician to prove that the dark areas were dross and not just a surface blemish as the vendor insisted.  The dross extended for much of the flange circumference.

Figure 3: Mill flange defect – offered for inspection after machining completed and ready for rubber lining.

So, What is the Point?

It is clear from these examples that not much experience in QC is required at all to detect significant flaws in castings.  The concern here is that in these cases and in many others, it was the owner’s consultants who found the faults, not the vendor’s inspectors.  In light of these examples, consider then Svalbonas’ complaint that owner’s structural consultants are not automatically expert in QC.  Whilst Svalbonas’ complaint is factual, i.e., structural expertise does not necessarily confer QC expertise, it is evident that almost any mechanical or structural consultant worth his salt would have been able to find these defects.

So, there are plenty of examples of substandard components offered by vendors to owners for acceptance and many times, the owner will not only find a fault but they will save their bacon and the vendor’s bacon (in terms of warranty).  Furthermore, we do not dispute that there are instances where owner’s consultants do not offer value in the evaluation process.  No-one is always right, not mill vendors and not consultants but as long as vendors serve up appalling components such as those in Figures 1 to 3, there is no option for the owner but to invest in their own surveillance team.

A Deeper Look at Castings:

In [1] and [2], Svalbonas makes the following points (in addition to his customary swipe at consultants):

1. The evaluation of sub-surface flaws using fracture mechanics is too conservative;
2. The evaluation of dross as surface cracks is too conservative.  He discusses “mild dross” and “more severe dross” as having certain material property reduction factors relative to pristine castings;
3. Separate subsurface flaws in close proximity will not amalgamate under the “stress levels used by Metso”.
4. It is not necessary to remove all dross from a casting or to scrap a casting that has dross.
5. The standard UT used to size sub-surface flaws is too conservative and there are phased-array devices available that can provide more detailed description of flaws to delineate individual pit clusters.

Well, we agree with the thrust of most of the claims in [1] and [2] but:

1. Whilst the evaluation of sub-surface flaws is too conservative, Svalbonas does not provide an alternative.  We suspect that his idea is not to evaluate sub-surface flaws at all, but it is unlikely that owners would accept this without an independent study being done.  Whilst Metso apparently has done some testing, the data has plenty of holes in it.  A further difficulty is that Metso, FLS and Outotec evaluate flaws differently from one another.  The “potato” shape described in [2] is not smoothed by other mill vendors; the complete stress range is considered by some vendors and not others; and stress intensity thresholds used by the various vendors to assess flaws vary by up to 50%.  Since there is no consensus between vendors, it is unlikely that there will be consensus on this matter between owners and vendors.
2. We do not believe that dross is a crack as in Item 2.  It is possible that a fatigue strength reduction could be appropriate but again there are problems.  Noting that “dross” means “worthless or rubbish”, how can we define the properties of Svalbonas’ “mild dross” and “severe dross”?  The material properties of pristine SG Iron can vary dramatically between foundries, will the properties of “mild” and “severe” dross also be different in all foundries?  Can the microstructure of dross be used to determine the fatigue strength reduction factor?  Can non-destructive testing methods determine what type of dross is present?  We are not rubbishing Svalbonas’ argument here, these are questions that he has not answered.
3. We agree that separate clusters of pits would not necessarily coalesce under load.  However, there is no point delving into this further if there is no way of detecting these separate pit clusters.
4. We agree again that the presence of dross should not result in automatic rejection.  However, in the photos of castings with dross in [1], the dross is located on the end of the trunnion that fits into the head and there is very little stress there.  That is why we didn’t complain about the casting when we were asked to review it by the owner.  What happens though when the dross is in a highly stressed area?  Is it reasonable to expect an owner to accept castings like those in Figures 1 to 3?  The owners we worked for thought not.
5. It is all good to say that there are phased array devices that can assess flaws more accurately than normal UT.  However, this equipment is not offered by Metso on all  casting flaw evaluations even when the casting is going to be rejected.  On a recent project, the owner all but insisted that such a device be used to assess a flaw that was sufficiently large to reject a casting yet Metso would not use it and gave no reason why.  Sadly, the casting was rejected.  There is not much point advertising this equipment if it not available for use.

The European standard on ultrasonic testing of SG iron castings, EN 12680-3: 2011, does not help here either because, for Severity Level 2 castings, dross is limited to 15% of the wall thickness and so the castings above would be rejected and the maximum size of subsurface flaws within 30 mm of the casting surface is only 1000 sq. mm (less than 2 sq. in.).  Severity level 2 may be too sever though.  In any case, this EN standard does not provide guidance on what to do if the flaws are larger then the specified acceptable limits.

Conclusion

When castings are as poor as those shown in Figures 1 to 3 above, no QC qualifications are required to know that they should be not be accepted without remedial work or in some cases, rejected outright. We cannot say that no-one wants to reject a casting without good reason but we do not know of anyone who does.  Instead of complaining about consultants, their qualifications and their experience, it would be more advantageous for mill vendors and the mining industry to put their hands in their pockets and fund some independent research that can be published openly and subjected to peer review.  Both vendors and owners should be involved so that bias is not an issue.

We agree with most of the issues raised by Svalbonas on the overly conservative nature of casting flaw evaluation but he has not presented any alternative that can be applied to castings from Metso and other vendors.  However, credit must be given to Svalbonas as it is clear that he has done a lot of research.  There is no evidence of such research being done by other trunnion-mounted mill vendors.

In essence, the point of view one takes on casting flaws  depends on perceived risk.  At present, there is no objective consensus on the issues described in [1, 2] and discussed in the bullet points above because not enough research has been published in the open literature.  Owners therefore are more conservative than vendors because the owners will face consequential damages if failure occurs whereas vendors will not.

Finally, it is worth noting that it is pretty easy for any competent mechanical or structural engineer to learn all that is necessary on the QA/QC of mills castings in a relatively short time.  It is not a matter of expertise in using test equipment.  What is important is how the results from an inspection are used to assess the casting flaws and yet there is no consensus or uniformity in non-conformance assessment within and between vendor groups and between projects.  This is not good for generating confidence in owner’s teams.  Svalbonas has proposed that an international standard should be developed for mill design [3].  This seems like a good idea that should be supported by industry.  It would certainly go a long way towards stopping disputes.

References:

[1] Svalbonas, V., et al., 2009. Mill head castings – educating opinions. SME Annual Meeting, Denver CO.

[2] Svalbonas, V., et al., 2006. Mill castings design – Experience vs. theory.
SAG2006 Conference, Vancouver, B. C.

[3]Svalbonas, V., et al., 2001. Fitness-for-purpose: should you buy one grinding mill for the price of two? SME Annual Meeting, Denver CO.

[4] Riposan, I., et. al., 2008. Surface graphite degeneration in ductile iron castings for resin molds. Tsinghua Science and Technology, v. 13, n. 2.

1 – Past Failures and Difficulties:

The most notable failures and difficulties in mill foundations have been caused by gear drives.  Some examples we have worked on are:

Mt Leyshon
The three foundations at Mt Leyshon all failed within a short time.  The failures were catastrophic, i.e., the cracks in the piers completely failed and there was no way that the mills could be operated.  In c. 1992, one of our staff was asked to determine why these foundations failed and to model a proposed fix.  The failure was traced to the gear forces as these matched one of the natural frequencies of the foundation piers.  Each mill was down for nearly a month while the repair was made and the concrete allowed to cure.

North Parkes
The next foundation we worked on was at North Parkes.  There, the foundations on the original plant would transient manner.  We were asked to design the foundation for the new plant from the point of view of vibration.  A block between the mill bearing piers was required to avoid vibration that would have occurred due to the combination of raft dimension, mill frequencies and soil properties.  There were no issues with vibration on these new foundations.  This work then led on to the Cadia project and many others (thanks to Geoff Cullen and Alan Boughey).

Fimiston (KCGM)
Another interesting foundation we studied is the Fimiston SAG mill foundation.  The gears on the mill on this foundation were very difficult to align and the alignment and vibration of the system was worse when the mill was rotated in one direction rather than the other.  We found that the direction of rotation would result in a change in the vibration levels at the pinion piers of up to 7 times depending on the level of damping in the structure (from 5 mm/s to 35 mm/s).  This was the first analysis that explained why the direction of mill rotation affected on vibration levels.

Fimiston mill foundations –
Mill and soil not shown for clarity

Red Dog Mine
One of the more interesting projects that we have been involved with was the Red Dog Mine SAG mill foundations.  The mills are supported on steel supports rather than concrete.  The mills operated for many years without issue and then began to vibrate excessively.  A consultant/mill-designer investigated the problem and recommended substantial stiffening to the steelwork apparently to overcome “softening” of the steelwork due to long-term deterioration.  We thought that it was odd that the supports would suddenly start to vibrate and recommended that the pinions be changed out.  This solved the problem and saved the owner a lot of money.  So again, gears and pinions were the main cause of vibration.

Gearless Mill Drive Foundations:
We have never encountered a foundation vibration problem with operating gearless driven mills, although there may be examples of this about.  The main problems with gearless driven mills is stiffness and strength; there are several examples of cracked foundations at the supports for ring motors.

We have stated in a previous post that the stiffness of ring motors and therefore the resonant frequency of the stator is dependent on the stiffness of the foundation supports; see: http://www.eand.com.au/?p=168.  Inclusion of some small details in a foundation of standard configuration (mill and motor piers connected) can produce a change in the natural frequency of the motor of about 5% and this is always worth doing as it does not result in a significant cost.  Once the piers have been sized and the details developed so the motor does not vibrate, then the system assessment reduces to ensuring that the foundation does not vibrate due to mill forces, i.e., avoiding global vibration including the soil-structure interaction.  This is the same task encountered when designing a gear driven mill foundation once the gear forces have been managed.

In terms of actual vibration problems with ring motors, we know of the Cadia vibration issues only.  There, we proved conclusively that the vibrations were not related to the design of the foundations.

2 – Difference between Geared and Gearless Mill Foundation Design

There is a key difference in system analysis of geared and gearless mill foundations.

1. In gearless motor foundations, one of the main tasks is to ensure that the foundation prevents the motor from deflecting or vibrating excessively.  If this is done properly, then the motor and the motor piers will effectively behave as static structures.  The remainder of the design analysis then involves the assessment of the other dynamic forces, primarily from the mill.
2. In gear driven mills, there will always be vibration from the drive and the mill.  This cannot be eliminated.  The aim of the analysis is to ensure that (a) the foundation does not resonate or (b) that there is enough stiffness and damping in the foundation and the sub-surface soil/rock to ensure that the vibration does not result in failure if resonant frequencies cannot be avoided.

So, for one mill type, the system analysis is mostly about making sure the equipment (the ring motor) will not vibrate or deflect whereas for the other, the primary task is to make sure the foundations will not vibrate.

3 – How to Design a Mill Foundation

The steps are:

1. Understand the loads – thermal, accident (short-circuit), gear loads, mill loads, liner frequencies.  You will need to do some measurements to get these loads;
2. Get the soil properties – Note that the soil properties will be approximate and you will need to perform a sensitivity analysis for the soil properties.  The extent of the soil/rock model under the foundations should be an order of magnitude greater than the foundation characteristic dimension in each of three directions;
3. For gearless drives, get the motor vendor not the mill vendor to define the required stiffness for their motor to behave properly;
4. Create a model that includes the mill, ring motor, pinions, gears, concrete and soil as necessary;
5. Assess the critical performance parameters, viz., deflection of motor and piers; vibration levels; hold-down forces, motor behaviour, etc., for all conditions;
6. Then, Bob’s your uncle and you are done.

None of the “answers” you get will be “right” because you won’t know the forces to within 10 or 20% and you certainly have no way of being sure the soil properties are right.  So, it will almost always be necessary to iterate to get the “right” design.  The foundations may vibrate under certain soil property combinations and this will require some changes by the concrete designers.  The ring motor may overload the hold-down bolts under accident conditions needing changes to bolt materials or increases in size.  In the mid-2000s, it was sometimes necessary to increase the stiffness for ring motor pier connections to avoid stator vibration due to very soft soils.  This is less likely to be a problem now but the possibility should always be assessed.  If motor diameters increase or power draws increase, the present design of stator frames may well require modification as a result of these system investigations.

There is no point just doing a modal analysis for gearless or gearless drives.  These would not have detected any of the GMD issues at Antamina, Collahuasi, Cadia or Freeport.  A modal analysis is just the first step in the overall dynamic assessment.  Separating the forcing and resonant frequencies by X% will not be good enough as there are too many frequencies to separate.

The length of the foundations is very important in determining behaviour.  This means that SAG mill foundations behave very differently from ball mill foundations.

4 – Mill Designer Stiffness Criteria

One of the mill designers performs a calculation to determine the “target stiffness” of the foundations given the mill and ring motor stiffnesses.  We have noticed that this ends up at about 10 MN/mm for large drives.  We have pointed in another post that the Freeport 38 ft installation has a stiffness of about 1.2 MN/mm and the Cadia 40 ft stator has a stiffness of about 2 MN/mm.  It is evident that 10 MN/mm is overkill.

5 – Mill Vendors being Difficult

We find that some US mill vendors will not provide drawings to enable the inclusion of the mill and bearings in a system model.  They complain about proprietary information and the danger of the loss of their intellectual property.   We are constantly amazed by this argument and it is up to the Owner to overcome this so that the system design can be completed in a reasonable time-frame.  All that is required is enough information to model the overall dimensions of the mill; mill component weights and stiffnesses; liner details; gear dimensions, PCDs, pressure and helix angles; bearing housing details; and bearing stiffnesses. In other words, what is needed is a set of engineering drawings with dimensions and no tolerances.  The secret IP of a good mill design is in the detail and tolerancing, not the overall dimensions, so the only thing that withholding this does is delay the project and increase risk.
.

6 – Some Myths

1. Multiple mills on a single foundation cannot be modelled (from a mill vendor’s design consultants).  Wrong.  We have modelled Lumwana, Boddington and others successfully (see example of a multiple mill model below).
2. The soil is not important and the foundation model can be fixed under its base for analysis.  Again wrong.  The El Teniente SAG mill foundations had to be stiffened as did the stator frame to ensure there would be no vibration because the soil properties were very poor.
3. Foundations cannot be built on piles.  Wrong again.  There are successful examples of piled mill foundations.
4. The mill vendor should be responsible for the system analysis.  Wrong.  A mill vendor is experienced in mill design and generally knows nothing about ring motors or foundations.
5. A geared mill does not require a torsional analysis.  Wrong for several reasons but for this post, the torsional harmonics can be greater than the dynamic loads of tooth mesh, mill and pinion rotation and liner loads.
6. Modal analysis is sufficient for assessment of mill systems. Not so.  There are many mill systems that operate at resonance.  There are too many forcing frequencies to consider in order to avoid resonance.  Furthermore, the vibration of a ring motor is not resonant.  It is sub-harmonic.  Modal analysis is not sufficient.
7. Motor vendors understand the hold-down forces.  This is not universally correct.  On vendor does not use a dynamic analysis to determine peak hold-down forces during a short circuit whereas the other does.  The one that doesn’t is wrong.

Multiple mill foundation. Soil not shown for clarity.

SUMMARY

Most of the major problems related to foundation design are from gear driven mills, not gearless mills. System analyses are useful for both types of mill.

The main aim of a system stiffness analysis is different for geared and gearless mills.  In the former, the aim is to stop excessive vibration in the foundations; in the latter, it is to make sure that the ring motor acts as a static structure.

Mill design consultants produce stiffness targets that are too conservative.  The Cadia and Freeport foundations are excellent examples of what can be done without going to 3 and 4 m wide motor piers.

This is an update of work presented at the SAG 2001 conference on the Cadia gears [1].  Both gears and pinions on twin ball mills failed almost simultaneously.  Our 2001 paper showed that the location of the failures is explained by the shape of the girth gear blanks.  Fresko et. al. [2] shows that T-gears are in fact better than Y-gears in terms of load distribution and that the diameter of the pinion in a gear set can result in excessively high loads at the ends of the pinions.  A more refined analysis of the Cadia gear sets performed by EAnD in 2006/7 shows that the load on the Cadia gears was very poorly distributed and this poor load distribution is even worse than that estimated in [2].  The location of the failures at Cadia can only be explained by thermal ratcheting.   Regardless of the differences in all work done, the conclusion is the same, the Cadia gear set failed due to design and not anything to do with operation of the mills, i.e., lubrication choices.

DISCUSSION

In 2001, we published a paper [1] that showed that the scoring of the Cadia girth gears was related to the thermal ratcheting of the gear set.  In summary, the analysis showed that the non-uniform shape of the girth gear resulted in differential thermal expansion, i.e., more expansion on one side of the gear vs the other.  This expansion explained why one of the girth gears at Cadia failed at the drive end and the other failed at the non-drive end.  The expansion is shown in Figure 1.  The work in [1] was limited by the size of the model we were able to create to simulate the failure.

Figure 1:  Differential expansion in a girth gear

In 2004, Fresko et. al. [2] presented a paper that showed that a T-shaped gear resulted in a load distribution across the tooth that was about 7% more uniform than the “Y-shaped” gear in Figure 1.  This paper presented a more advanced model of gear sets than that used in [1] in that both the pinion and the gear were modelled together to estimate the load on the tooth flanks.  The limitation of the paper is shown in Figure 2.  Whilst the helix of the teeth was modelled, the contact was idealised as acting along the pitch line of the gear face simultaneously and only one tooth was in contact.  This is effectively considering the gear and pinions as being spur gears but still, this is an improvement over the model in [1].

The gear tooth loading profiles for T and Y-shaped gears obtained in [2] are shown in Figure 3.  Interestingly, the “alignment” of the pinion for the T-gear seems to have been chosen so that it results in a misaligned pinion over the rib thereby resulting in a higher peak load at the non-drive end of the pinion.  This is an odd choice if the T and Y-gears are to be compared equally but nevertheless, the load factor for the T-gear is 7% (1.91 vs 2.04) better then the Y-gear.  It is clear that this could improved by a further 5% if the pinion were “aligned” over the rib rather than over the window.

A further conclusion from [2] is that the pinion diameter is more important that gear blank profile (Y or T).  So, according to [2], the failure of those gears was caused by a design fault, i.e., the diameter of the pinions in the Cadia ball mills are too small.

Figure 2:  Model of gear contact in [2] – similar to a spur gear, not a helical gear

Figure 3:  Load distribution from [2] for T and Y shaped gears

In 2006, we undertook a further analysis of the Cadia gear set because PCs had reached a level that allowed detailed modelling of the gear, the pinion and the actual tooth contact line.  The model of the gears is shown in Figure 4(a) and the typical calculated contact pattern for both of the Cadia gears (there are two mills at Cadia) is shown in Figure 4(b).  Note the following:

1. There is more than one load line in the load plot because there is more than one tooth in contact.
2. The tooth contact line was determined by the intersection of the two involutes of the mating teeth, i.e., the contact occurs along a line that is inclined to the pitch line.
3. There is almost no load taken at the centre of the gears due to the relative stiffness of the mating gears.  This load variation is much greater than that obtained using the approximate methods in [2] for a Y gear.

The contact load patter in Figure 4(b) implies that the Cadia gears should have both failed at the drive end of the pinions.  However, they did not.  One failed at the drive end and the other failed at the non-drive end.  The only explanation for this given the alignment process that was undertaken prior to the failure is that the Y gear expanded preferentially at the cope end and that the cope ends of the gears match the locations of the scoring.  Indeed, adding heat to the gear simulates this effect.

The damage to the girth gear was on both the loaded and unloaded flanks.  Site inspections (see Figure 5) concluded that the backlash had been entirely taken up by the relative expansion of the gear to the pinion.  This is consistent with thermal ratcheting.

(a) Gear and pinion model for Cadia gears

(b) Typical load distribution on Cadia gear teeth

Figure 4:  Model of Cadia gears with correct contact and resultant load distribution

Figure 5: Damage to girth gear
(Note: The damage to the loaded flank can be seen on the under-side of the top tooth; the damage to the unloaded flank can be seen on the top sides of the three lower teeth.)

CONCLUSION

The gears and pinions on the Cadia ball mill gears failed due to a design fault.  We agree with [2] that a larger pinion may have avoided the gear scoring, however, the location of the damage can only be explained by thermal ratcheting resulting in differential expansion.  The damage to both the loaded and unloaded flanks of the girth gear are consistent with the gears almost jamming.

References:

[1] Meimaris, C., Duncan, M. D., Cox, L., 2001. Failure analysis of ball mill gears, SAG 2001 Conference, Vancouver.

[2] Fresko et. al. 2004. The use of finite element analysis in the understanding of alignment and load distribution in large grinding mill gear and pinion sets.  SME Annual Meeting, Denver, CO.

Svalbonas’ paper paints the state-of-the-art in ring motor design as being in a parlous and undeveloped state in comparison with that of mills. Is this is reasonable depiction of the two machines?  The answer is “No”. We have recently undertaken failure analyses of six mills built using all the modern design methods described [1], so mills still do fail. The structural analysis of some of the largest mills in the world is primitive when compared with the design technology available in general industry today. There is no uniformity in design approaches between any of the mill vendors.  Metso, FLS and Outotec mills are designed using different detailing and allowable stress ranges for welds and there are major differences in the way these companies assess the structural integrity of SG iron castings. Furthermore, the design verification of mills by strain gauging has not always been performed properly as in [2] and elsewhere, yet reliance on this strain gauging is the cornerstone of Svalbonas’ viewpoint on ring motor validation.

1 – Ring Motor and Mill Complexities

We have been fortunate to undertake detailed mechanical and structural design reviews and audits of all mill types from the five major mill vendors and of ring motors from both the gearless drive vendors.  We have also been hired by mine owners to determine the causes of most of the major modern mill and ring motor failures that have occurred since 1998. This gives us a unique advantage in comparing the state-of-the-art in design for mills and ring motors. No mill or motor vendor, mill design company or independent consultant can truthfully claim this advantage.  So let’s look at the observations in Svalbonas’ paper on gearless drives in [1].

The structural design of mills is much simpler than the structural design of ring motors. Most of the complexity in a mill is related to bearing design. However, Svalbonas restricts himself to structural design in his comparison of mills and ring motors in [1]. The rotating element (the shell, heads and trunnions) of a mill is a slowly rotating tube. Its structural design is often analysed using a simple, linear, two-dimensional finite element model and there is no need to undertake a dynamic analysis of the mill structure.  A ring motor on the other hand is much more complex, be it an ABB or a Siemens design. The stiffness of a stator relies not only on its steel frame but also the core that the frame is designed to support. The core is not a homogeneous solid; it is a lamination of approximately 2000 sheets of steel, each about 0.5 mm thick and the stiffness of the lamination is dependent on the compression applied by lamination clamping bolts and plates. The attachment of the core to the frame is via one hundred or so keys that are either welded or wedged to hundreds of brackets. These brackets are welded to diaphragms in the frame and the diaphragms are welded to the three or five main rings of the frame.  The completed frame cannot support the electromagnetic loads in the core by itself as it is not stiff enough.  The final step in generating sufficient stiffness in the stator is achieved by the over-constraint in the supports of the frame on its foundations. The stator is fixed on both sides of the frame or at the bottom. There is no sliding at the foundation supports to prevent the generation of large thermal expansion loads in the frame and the foundation during operation. These thermal loads are only limited by the cooling system of the stator. A mill is much simpler. It is not over-constrained on its supports as it has fixed and floating bearings and there are no thermal loads to contend with. Even the rotor poles of a ring motor are more complex than a mill structure. They too are dependent on the steel laminations for stiffness and they are subject to highly non-linear forces as is the stator. The loading in a mill is simple in comparison as stated in [1].

The design task faced by the motor designer, even if limited as in [1] to “structural design” (which is clearly mechanical), is vastly more difficult than the task of the mill rotating element designer. There are more components to design. The stator and rotor components are subject to complex loading including non-linear electromagnetic loads in the air-gap and the thermal loads from the electrical losses. The electromagnetic loads generate high mean stresses in stator and rotor components as well as fatigue loading. The motor must be designed for accident conditions such as those produced by short circuits. These accident loads result in highly transient dynamic loads in which the rotor pull is released but the torque increases by roughly an order of magnitude. In contrast, mill vendors do not have to design for an accident condition (e.g., a dropped charge).

In summary, the rotating element of a mill is a simple structure that is intrinsically stiff and it is designed to accommodate well-defined cyclic loads that occur once per revolution. A ring motor is a complex machine that relies on its steel frames, the cores they support and the foundations (in the case of the stator frames) to achieve sufficient stiffness and is subject to static (thermal), cyclically dynamic and high-amplitude transient loads. Mills and motors are chalk and cheese in terms of complexity.  This greater complexity of ring motors is considered only superficially in [1].

2 – The Failures Mentioned in [1]

It is interesting to consider the choices of subject in the paper and the way they are presented. We note that we were responsible for the failure analyses of all the failures related to ring motors mentioned in [1] and most of the design changes made to avoid these failures in subsequent motors. Svalbonas was not involved in these failure analyses nor has he contributed to the remedial designs or design changes to subsequent motors that arose therefrom. We also undertook a failure analysis of the KCGM mill and have analysed forged fabricated gears, both of which are mentioned in [1].

2.1 – KCGM mill and forged-fabricated gears

The review of the KCGM mill and mention of fabricated gears in [1] offer no benefit in the understanding of the ring motor failures or their design.

The cracks in the KCGM mill shell are primarily due to the installation of the mill. All the joints in the mill were lined up so that the head and shell joints all pried open resulting in very localised stress concentrations in the joints and eventually cracks in the welds. The comment in [1] that the “mill continues to run because we know that the local detail failure poses no immediate danger to the mill [due to] FEA studies and a history of mill strain-gaging instrumentation” is not true. The reason why we “know” that the mill has not failed is precisely because the mill has not failed. It is hindsight that tells us this, not strain gauging or FEA. Otherwise, the Owner would not have spent money investigating methods such as add-on brackets and doubler plates to ensure the cracks do not propagate.

In contrast to the cracks in the KCGM mill that seem dormant according to [1], consider the Candelaria SAG mill failure described by Yanez and Tapia in [3].  The SAG mill cracked through the flange and into the shell and this threatened the integrity of the mill.  The crack repair required removal of balls and liners and half the head to enable access to the the crack.  This was as substantial an undertaking as the repairs to the Cadia rotor poles and required more time to complete.  Again, FEA and strain gauging were not necessary to assess the danger that the crack posed to the mill because it was sufficient to observe that the crack was growing.  FEA and strain gauging in isolation do not make components “safe”;  good overall design does.

The comparison of the details in the gussets in the Cadia rotor poles with the scalloped gussets in fabricated gears in [1] is misleading. The root cause of the rotor pole failures is unrelated to the shape of gussets. Furthermore, the loading of gears and rotor poles is completely different.  There is no fundamental reason to believe that fabricated gears with gussets will result in premature failure.

2.2 – The Cadia rotor pole failure

EAnD were engaged by the mine owners to determine the technical (proximal) and root causes of the stator vibrations and the rotor pole failures in the Cadia motor by Newcrest Mining, the owner. In both cases, we undertook dynamic measurements of the equipment followed by complete finite element analyses of the ring motors to determine the causes of failure. The changes made to the motors were recommended by EAnD and these have been adopted by Siemens in the all their new motors. The reviews were published in various papers [4, 5, 6, 7].   The discussion we present on ring motors is based on facts and not supposition.

2.2.1 – History of the Siemens ring motor design

The failure of the welded ribs in the Siemens large diameter motors was first observed at Cadia. We measured the stress range at the toe of the weld using strain gauges as shown in Figure 1 and found that it was slightly greater than the allowable stress range in fatigue design codes. The strain measurements showed that stiffness of the rotor pole laminations was substantially less than the value used in the design. We found a similar result in the investigation of vibrations in the stator [7]. The proximal or direct cause of the rotor pole failures is an  error in estimating material properties. It is not due the welding design, the use of gussets, attempts to avoid intersecting welds, weld accessibility, or welding stop-starts as stated in [1]. The choice in the repair of the poles to change the weld design from partial to full penetration welds and to grind the weld toes was made because it was more practical to upgrade the welds in a site repair than to increase the stiffness of the laminations.

Whilst the direct cause of the Cadia vibration problems and the rotor pole cracks is related to estimation of material properties in laminations, it does not explain why the designers did not catch these errors during the design phase of the motor, i.e., it is not the root cause of the failure.  The origins of the failure can be traced back to c. 1994 when PT Freeport Indonesia ordered its first gearless drive for their 34 ft SAG mill.  The motor had to fit through a tunnel on the road to the site. This necessitated a change in the frame design from that used in older motors to allow clearance in the tunnel [8]. The next motor for Freeport was for a 38 ft mill and this was ordered in April 1996. Again, the components of the motor had to fit through the tunnel and the size of the frame was determined by this requirement. The depth of the rotor poles was determined by the frame diameter. The Cadia motor was ordered in July 1996. The design of the Cadia motor followed the evolution of the Freeport motors in that the frame was more slender than older designs.  The frame design was a compromise between frame size and weight and foundation size and geometry [6].  The Cadia frame reached the limit of frame design slenderness and required some modification but this did not result in major downtime losses [7].  However, the compromises made during design to minimise the footprint of the motor [6, 8] resulted in shallow rotor poles, as the depth of the poles is determined by the outer diameter of the frame. Whilst the direct cause of rib weld failures in the rotor poles is stress-related (via the stiffness of the laminations), the decisions made by plant designers in the decade prior to the failure influenced the choices made by the ring motor designers to the extent that the rotor poles were too shallow and thus failed.  The designers did not recognise that complying with these external requirements in terms of motor weight and size resulted in the stator and rotor frames becoming so slender/shallow that the design of the motor entered a new regime in which the stiffnesses of the laminations (cores) became critical to the structural integrity of the stators and the rotors themselves.  The failure to recognise that these external factors were changing the fundamental behaviour of the motors components was the root cause of the rotor failures and the initial stator vibration.  The design of the Siemens motors has evolved substantially in the 16 years following the Cadia motor commissioning.

Figure 1 – Strain gauging of rotor poles

Strain gauging of rotor poles at Freeport most likely would have provided the correct material properties for the laminations in the rotor poles as stated in [1] but even with these properties, measured at Cadia, the designers struggled to get past the slenderness issue. The “wings” added to the Cadia rotor poles definitely were not deemed advisable after more FEA of the poles as stated in [1], at least by EAnD and the owner. This was a belt and braces addition by Siemens that we knew would not provide much benefit in terms of strength but the owner did not argue about their inclusion of the wings as they were not detrimental.  Unfortunately, the wings were included in some motors ordered after the Cadia failure but this design has since been discontinued [11].

2.2.2 Legacy of the Cadia and Freeport motors

If looked at properly, the Cadia motor should be considered as one of the best designs of its type.  Whilst it did have two problems, neither resulted in substantial downtime.  The vibration issue was controlled by site maintenance and the vendor without any significant loss of availability until the strongback stiffener could be applied.  The cracked rotor poles did not result in any damage to the windings again due to good maintenance.  The motor was designed just outside the envelope of trouble free operation and the research undertaken to resolve the issues defined the envelope limit.  Knowing these limits has benefited the entire mining industry.  The overall system design including the foundations is one of the most cost-effective in the world.  It would be of benefit to mine owners to study the designs of the Cadia and Freeport motors and their foundations before embarking on new projects.  Foundations are now becoming too large and the stiffness targets for these foundations are excessive given that stator frames are now stiffer than the Freeport and Cadia designs.  A lot of money is being wasted in concrete.  The Cadia and Freeport foundations have stiffness levels of about 2 MN/m whereas newer installations are being designed with stiffness targets of 7 to 10 MN/m to support stators that are substantially stiffer than the Cadia and Freeport stators.

2.3 – Collahuasi hanger plate failures

The design of the hanger plates was looked at in depth during the failure analysis we performed for Collahuasi in 2006/7. Svalbonas’ statement in [1] that the failure of the hanger plates is not related to diameter is incorrect.

The failure of the Collahuasi motor was catastrophic and it occurred within months of commissioning.   The Collahuasi motor was for a 40 ft SAG mill. The Antamina 38 ft SAG mill was also driven by an ABB motor and the hanger plates had not failed on that machine after several years of operation. Similarly, the Sossego motor for the 38 ft SAG mill had not failed.  It too had an ABB motor and it had operated longer than the Collahuasi motor. Inspections of the Sossego motor indicated that the hanger plates had sharp corners in many of the hanger plates just as the Collahuasi motor had and that the magnetic pull per hanger plate is approximately the same as for the Collahuasi motor. Both motors driving the Telfer 36 ft SAG mill have hanger plates with sharp corners and had operated longer than the Collahuasi without failure.  So why did the Collahuasi fail so quickly when the others did not?

Firstly, it is important to note that the failures at the corners of the hanger plates were not simply due to “stress concentrations” as stated in [1].  The re-entrant corners of the hanger plate cannot be analysed using stress concentration methods because both the design and manufactured corner radii were small. Below a certain value of corner radius, maximum stress adjacent to the corner radius does not increase. This is because the slot for the keybar in the hanger plate acts like a crack rather than a stress concentration. The hanger plates would have failed even if they had been made with the 3 mm radii called for in the design. So, the hanger plates in all the ABB ring motors were effectively “pre-cracked” prior to the motors being commissioned.  Our recommendation to Alstom (the ring motor builder) to use a 10 mm radius for the hanger plates in the replacement Collahuasi stator was made to avoid the fracture mechanics regime in the hanger plate design.

Secondly, the diameter of the motor comes into play. The stiffness of the stator core is a function of diameter, thickness and depth. Stiffness is proportional to the cube of the core diameter, thus a small change in diameter can result in a large change in core stiffness. In smaller motors, the core can be self-supporting to a certain extent, i.e., it can act as a ring that does not transfer all of the electromagnetic loads to the frame. In fact, some of the vendor’s senior electrical engineers believed that the ring was completely self-supporting.  Now, as the diameter of the motor increases, the global bending stiffness of the core decreases and more of the weight of the core and the electromagnetic forces in the air-gap are transferred to the frame of the motor through keybars and hanger plates. So, the Collahuasi motor failed much more quickly than other smaller motors primarily because of the diameter of the motor itself.

2.3.1 Lessons from Collahuasi

The Collahuasi motor resulted in extensive downtime for the plant and substantial losses.  It is simple to point to the individual design faults in hanger plates, insulation, keybar retention and lamination failures for this motor and subsequent ABB motors as the cause of the failures but this is like not seeing the forest for the trees.  The Collahuasi failure/s were cause by a lack of design oversight.  This was specifically highlighted as a potential risk prior to the start of work on the project after an inspection/assessment of the ring motor factory.  However, there is a strong desire in owners and even more so, in engineering companies, to not be involved in detailed design reviews for fear of attracting risk. We suspect that all the design reviews of all the mills and ring motors built since the Collahuasi motor was ordered would have cost less than the losses suffered at Collahuasi.  Not reviewing the design just does not make sense.  Even if a design review only has a slim chance of detecting a fault because of the complexity of the design issue, not performing a design review ensures there is no chance of finding it.

3 – Validation by Strain Gauging in Ring Motors and Mills

Svalbonas states that the failures in ring motors were not predicted by previous finite element analyses because insufficient strain gauging of motors had been undertaken on ring motors in the past. He contrasts this with the mills where strain gauging has been undertaken to validate design models.  Reality is not so simple.

3.1 Design validation in ring motors

The strain gauging of the Cadia rotor poles (Fig. 1) was useful in providing an accurate estimate of the stiffness of the pole laminations.  Lower the lamination stiffness resulted in higher the stresses in the welds.  However, the data alone was not sufficient to avoid failures of some motors designed after the Cadia pole repairs were completed because the paradigm under which these motors were designed had not shifted to incorporate the lower lamination properties. This shift in understanding has now happened (see Section 4).

Strain gauging would not have been a benefit for the Collahuasi failures. It is unlikely that any measurement would have prevented the hanger plate failures.

A typical hanger plate in the ABB motor is shown in Figure 2(a) and a failed hanger plate is shown in Figure 2(b).  It is clear from the geometry of the hanger plate, i.e., the sharp re-entrant corners, that the designers did not give much thought to fatigue and this is primarily due to the idea that the load in the hanger plates was not substantial.  The success of smaller diameter machines would have reinforced this concept.

(a) Typical hanger plate

(b) Cracked hanger plate
Figure 2:  Hanger plates in ABB motor

The strain gauging of hanger plates was undertaken by EAnD on the Collahuasi and other ABB motors as shown in Figure 3.  Several complete keybar rows were gauged. It was not possible to measure the total load supported by a hanger plate as the load path from the core to the keybars to the hanger plate is not the same in all motor locations or conditions.  At least 20% of the 500 plus hanger plates would need to be gauged to get a reasonable estimate of the load transfer between the core and the frame.  This is because there is a substantial hysteresis in the load in each hanger plate during a full load cycle.  The complexity of the motor precludes simple measurement techniques being used such as strain gauging to validate a design directly.  Real safety factors cannot be determined by measurement as claimed in [1].  The design process for the ABB motors is by necessity evolutionary, not deterministic.  What happened at Collahuasi was that the size of the motor extended well beyond the regime in which smaller motors operated.  The loads in the hanger plates became substantial and thus a change in design was needed.  Such a change in regime may be detected in design through analysis and reviews but it cannot be detected by instrumenting smaller machines as they do not operate in the new regime.

Figure 3:  Strain gauging of hanger plates

An in-house or external, detailed design review by a good designer who was not immersed in actually doing the design most likely would have detected the problems in the Cadia rotor poles.  There is no guarantee that the problem with the hanger plates in the Collahuasi hanger plates would be detected by such a review as it requires a paradigm shift but there was a possibility nevertheless.

3.2 Design validation in mills

Svalbonas says in [1] that ring motor designers must validate their designs against strain gauge measurements just as he has done in mills.  Whilst he mentions the difference in loading between mills and ring motors, he fails to consider the effects of the much greater complexity in construction of ring motors in his comments.  The stresses in the rotating element of a mill can be characterised with a relatively simple strain gauge measurement programme.  It would require of the order of 12 gauge sets. The complete characterisation of a ring motor requires strain gauging of rotor poles (about 8 gauge sets);  strain gauging of the various stator components (keybars, supporting plates and frame – 20 to 30 gauge sets), vibration measurements requiring about 30 channels, modal measurements for in-plane and out-of-plane modes, temperature measurements, flow measurements and determination of the foundation stiffness using a further 8 vibration channels. Whilst validation should be performed by the ring motor vendor, the task is much more complicated than applying a few gauge sets to measure some local stresses.

It would be unreasonable to dismiss the viewpoint of a mill designer on the design of ring motors just because the mill structure is relatively much simpler than the motor.  So, let us look at the design validation of mills.  Svalbonas [2] presents strain gauge measurements from a mill in a paper dated 1979.  This data was used to assess the accuracy of a finite element model of the mill.  The measured stresses in a particular direction are calculated in [2] by multiplying the strain in that direction by the elastic modulus.  This is incorrect as a biaxial stress state exists in the mill and the stresses are a function of both the radial and circumferential strains.  This error could be ignored if it were an isolated on but it is not.  We have found the same error in strain gauge measurements of larger mills taken more recently.  There is not much point validating design models against erroneous measurement data.

There is a bias in engineering that analysis is always doubted in the face of conflicting measurement data and this is evident in the interpretation of stress measurements in [9].  Stresses were measured at the weld toes of the can-to-can flanges in the Telfer ball mills by EAnD.  These stresses were substantially lower than the calculated stresses.  Svalbonas in [9] claims that it is “known fact” that a “standard” finite element analysis results in over-estimation of stresses at this location.  This claim is wrong as shown by analysis in [10] and by measurement on the Antamina mills.  The measured stress ranges at the toes of the welds at the can-to-can flange in the Antamina ball mills were slightly higher than those obtained from the “standard” finite element analysis that was undertaken during the design of the mill.  The difference between the measured stresses at Telfer and Antamina is that the liners in the Telfer ball mill had been wedged by ball scats and ore as shown in Figure 4 and thus the liners supported part of the load [10].  What this means to an owner is this.  If the owner accepts higher than allowable stresses at these joints on the basis of the erroneous “known fact” that FEA over-predicts actual stresses, the owner will accepting a mill design that relies partly on the liners for its structural integrity.

In future posts, we will present other design and design validation issues in mills.  However, the two examples above are sufficient to indicate that the state-of-the-art of design in mills is not substantially any different from than the state-of-the-art of design in ring motors.  There are still gaps in knowledge bases of the mill designers.

Figure 4:  Packing (wedging) of liners enabling axial load transfer

4 – Ring Motor Repairs

The design changes made to both Siemens and ABB ring motors after the Cadia and Collahuasi failures were more involved than the stiffening of the pole ribs and the increase of the hanger plate radii discussed in [1].  Again, this is due to the evolutionary design of these machines.

At the time that the Cadia rotor poles failed, the Siemens motor design had already gone through an evolutionary design change that resulted in stiffening of the stator frame [6, 7].  The rotor pole failures too required an evolution.  The description of the repairs to the Cadia poles by Svalbonas in [1] is not accurate.  The “wings” that were added to the poles were not “deemed advisable” by EAnD and the owner as explained in Section 2.1 above as they did not improve the strength of the welded joints in the ribs.  In subsequent motors, the wing design was used in Siemens motors except in cases where the motor was specified by EAnD.  We required that the rotor pole design be modified to eliminate the wings and to add two more rib sets to the pole resulting in one pair of ribs per pole.  This design was implemented at Antamina, RPM (Brazil) and Sino Iron.  Siemens have now adopted the four-pair rib design for all their new motors [11].

The changes made to the ABB hanger plates were part of a wide range of changes made to the design since 2006.  These changes address the failure at Collahuasi and other issues in the older machines through changes in insulation, lamination fixing and core expansion.  It will be some time before the effectiveness of these changes can be determined.

CONCLUSIONS

The failures of the rotor poles at Cadia and at other plants and the failure of the core support system at Collahuasi were not simply due to a lack of measurements on preceding ring motors as stated in [1]. Ring motor design is an evolution and at some critical size, the motors started to behave differently from earlier, smaller motors. As the motor diameters increased, the stator cores shed more load to the supporting structures. There would be little point in measuring the loads in a hanger plate in a smaller motor built in the past if that hanger plate does not support any substantial load.  Also, because of the load hysteresis, strain gauging of just a few hanger plates would not provide an accurate estimate of the applied loads.

Svalbonas’ paper portrays ring motor design and validation as being well behind the state-of-the-art in mills yet Svalbonas has had no involvement in ring motor design or in the analysis of any of the failures described in [1].  The recent spate of mill failures; errors in strain gauge data from mills; a general reticence by mill designers to investigate the discrepancies between measurement and analysis; and a wide divergence between the mill vendors in how they assess their designs and manufactured components does not support Svalbonas’ conclusion that mill designers have advanced further than ring motor vendors in determining the safety factor in their product.

The root causes of the failures at Collahuasi and Cadia were (a) faults in the vendors’ design culture and (b) indirectly to motor diameter.  Similar problems exist in the design cultures of mill manufacturers, as the absence of failures in the early 2000s has bred an air of complacency.  Unfortunately for all, this good run of mill operation without failures appears to have ended.

One thing that strain gauging and vibration testing of the Siemens and ABB ring motors has confirmed is that the designs are completely different both electrically and structurally.  This is poorly understood in the mining industry.

References:

[1] Svalbonas, V., 2008. Gearless motor failures – a mill designer’s viewpoint. MAPLA Conference, Chile.

[2] Svalbonas, V., 1979. Some considerations in computer structural analysis of large grinding mills. SME Annual Meeting, New Orleans, Louisiana.

[3] Yánez, G., Tapia, H., 2008. Reparacion de fisura de shell y giro de tapa de descarga del molino sag 36’x 15’ de minera candelaria. MAPLA Conference, Chile.

[5] Meimaris, C. and Lai, W. K. K. L., On the comparison between measured and calculated stresses in large SAG mills. Minerals Engineering 24 (2011)

[4] Meimaris, C., Lai, B., Price, B. F., Manchanda, S., 2006. How big is big? – revisited. In Proceedings of the SAG 2006 Conference, Vancouver.

[5] Meimaris, C. Higher Doctorate Dissertation, 2011

[6] Kummlee, H., Mienke, P., 2001. A mechatronic solution – Design and experience with large gearless mill drives. In Proceedings of the SAG 2001 Conference, Vancouver, Canada.

[7] Meimaris, C., Lai, W. K. K. L., Cox, L., 2001. Remedial design of the world’s largest SAG mill gearless drive. In Proceedings of the SAG 2001 Conference, Vancouver, Canada.

[8] Frank, W-D., Gearless drives – Experiences and new/future developments. In Proceedings of the SAG 1996 Conference, Vancouver, Canada.

[9] Orser, T., Svalbonas V., Van de Vijfeijken, M., 2011. Conga: the world’s first 42 foot diameter 28 mw gearless sag mill. In Proceedings of the SAG 2011 Conference, Vancouver, Canada.

[10] Meimaris, C. and Lai, W. K. K. L., Fatigue design of mills. Minerals Engineering, 30 (2012).

[11] Recent discussions with Siemens.

• Shell-mounted mills are not as reliable as trunnion-mounted mills because they rely on a T-welds that connect the end-plate to riding-ring (journal) (Svalbonas, [1, 2]).
• Trunnion-mounted mills are not as reliable as shell-mounted mills because they rely on large castings that are often found to have unacceptable flaws according to Knecht [3] and Leonard [4].
• Shell-mounted mills are not as stiff as trunnion-mounted mills according to Svalbonas [1, 2].
• Trunnion-mounted mills are not as stiff as shell-mounted mills according to Knecht [3] and Leonard [4].

It is evident that the claims of Svalbonas and those of Knecht and Leonard cannot both be correct.

The main problem with vendor publications regarding the choice between shell and trunnion-mounted mills is the potential for bias. Vendors often promote the benefits of their own design by directly contradicting the claims of the other vendors. For example, when Leonard was employed by Nordberg who built shell-mounted mills, he declared in [4] that shell-mounted mills represented a “revolution” in mill design that would “increasingly become the standard arrangement for large mills”. He justified his claim on the basis that shell-mounted mills did not require large castings that could delay project completion due to the presence of casting flaws; that they have greater stiffness that results in a “superior load distribution”, less maintenance downtime for gear alignment, and lower bending moments in the shell; and that they require smaller foundations without high piers and that this gives greater flexibility in drive arrangements. However, when he joined Metso in 2001 as its President, he did not transport the shell-mounted mill revolution. Metso continued to make trunnion-mounted mills. He did not eliminate castings from Metso mills and the stiffness of the Metso mills was not increased to improve load distribution or gear alignment. In fact, Svalbonas who was the Director of Engineering at Metso under Leonard states in [1] that the “much heralded elimination of castings [in [4]] actually lowers the overall safety of the mill structure” in shell-mounted mills. Svalbonas directly contradicts Leonard’s claim of higher stiffness in shell-mounted mills stating that both Leonard and Knecht [3] are wrong because they use simple beam theory to reach their conclusions on stiffness. Svalbonas raises several other potential difficulties in the design of shell-mounted mills in [1]:

1. It is more difficult to effectively seal the lubrication oil for the main support bearings in shell-mounted mills due to their large journal diameters;
2. The allowable stress ranges for welds in shell-mounted mills should be less than those in trunnion-mounted mills because the T-joint at the bearings has internal weld surfaces that may be exposed to the mill contents;
3. In segmented shell-mounted mills of the type built by Polysius, there is an unbolted section of the journal between the centre of the T-joint and the start of the inboard axial flange and this could open under load.
4. Shell-mounted mill designs have not been subjected to “analytic scrutiny and strain gauge study” to the extent that trunnion-mounted mills have been.

In the following sections, we will show that the disadvantages attributed to shell-mounted mills by trunnion-mounted mill manufacturers are either exaggerated or incorrect and that both mill types can be designed with the same or similar levels of safety.

1 – Stiffness

There is no particular problem with stiffness in grinding mills. Both mill types can be made sufficiently stiff. The issue appears to be a case of highlighting an attribute of a particular mill type that has no practical significance to the mill installation. Shell-mounted mills with end walls directly above the main bearings are generally stiffer than trunnion-mounted mills. We have checked this by comparing a 26 ft shell-mounted ball mill with a trunnion-mounted mill of equal grinding volume and same diameter.  Detailed finite element models of each mill were created. The number of nodes and elements in each model differ by less than 5%.  The elements used in the finite element mesh are parabolic with mid-side nodes.  Each mesh has four elements through the thickness of the shells and heads and the maximum element aspect ratio is less than 1.5. These are much more detailed models than any vendor uses in mill designs.  The same liner and charge loads is used in both models. The calculated overall stiffness for the shell mounted mill is 10.8 GN/m and 8.6 GN/m for the trunnion-mounted mill. Both of these stiffnesses are very large. Both are sufficient to enable proper operation of both gear and gearless drives. The stiffness argument is a distraction brought about by the competitiveness of mill vendors. It should be considered a peripheral issue when assessing mill types.

One stiffness related matter that is more important than the overall stiffness of mills is the local stiffness of the bearing surface. Svalbonas [4] argues that their trunnions are stiffer than the bearing surface of shell-mounted mills.  This is incorrect when applied to the standard shell-mounted mill built by Polysius, Outotec and FFE (in the past). These shell-mounted mills have an end-wall welded to the journal that stiffens the journal immensely. For the 26 ft mills in the previous paragraph, the maximum local deflection of the journal over a bearing pad is 100 times less in the shell-mounted mill than it is in the trunnion-mounted mill, i.e., the journal is 100 times stiffer in the shell-mounted mill.

So, shell-mounted mills with end plates directly above the bearings are stiffer than trunnion-mounted mills but both are stiff enough.

2 – The elimination of castings increases the risk in shell-mounted mills: Svalbonas [1]

Svalbonas’ argument is based on the premise that welded joints have lower safety factors than castings and since there are no castings in shell-mounted mills, they should represent more risk. This is incorrect since both mill types have welded joints and the internal surfaces of these welds can be exposed to the mill contents. If a volume basis of risk is used, then superficially, Svalbonas’ claim may be right (by weighting the lower risk castings and higher risk shells) provided we accept Svalbonas’ claim that castings have a safety factor of ~2 vs 1.3 for welds. However, trunnion-mounted mill vendors in North America neglect local stress concentrations in heads at the ends of split flanges and so the safety factor of 2 for castings claimed by Svalbonas is excessive. Weldments and castings will have similar factors of safety at the highest stress locations in a component. Furthermore, the observation that iron castings result in delays due to the presence of casting flaws in [4] is valid as experienced on many projects.

The arguments for or against the use of castings in mills are not a strong reason to favour or reject shell-mounted mills. This is supported by Svalbonas’ observation that the welds in both shell and trunnion-mounted mill types can be designed to the same level of risk [1].

3 – Effective sealing of the bearing surface

The larger the bearing surface (journal) the harder it is to seal. Our discussions with operators of large shell-mounted ball mills indicate that whilst the bearing seals leak, they do not a substantial maintenance problem. This may or may not be the case for larger diameter mills but the best people to contact regarding this issue are site maintenance personnel, not  trunnion-mounted mill vendors.  This issue is not a critical in terms of mill structural integrity.

4 – Risk from internal welds at T-joint

In [1], Svalbonas equates the internal weld at the T-joint with the internal corner welds in older trunnion-mounted mills.  These older mills failed prematurely more frequently than modern mills have (those with out-turned flanges). Svalbonas states that allowable stress ranges at internal weld surfaces of the T-joints must be lower than they are for external welds because they are they are subjected to corrosion, erosion, accidental damage and have a different ratio of tensile to compressive stresses than external welds. There are two problems with this argument, namely:

1. Modern trunnion-mounted mills also have welds with “internal” surfaces that are relatively highly stressed at the head to shell flanges and these surfaces can be subjected to corrosion, erosion and accidental damage as easily as T-joint welds internal in shell-mounted mills.
2. The comparison of the welds at T-joints made in European facilities today with those of older trunnion-mounted mills is inappropriate. The current best-practice for T-joints welds is that they are machined to a radius that is flush to adjoining parent plate surfaces (and below any toe defects). This results in much lower toe and peak stress ranges in the weld compared with the welds in many of the older trunnion-mounted mills that were only toe ground or not ground at all. The welding processes, consumables and parent metals used in these T-joints are of much higher quality than those used in older trunnion-mounted mills and the joints also undergo more stringent inspections than the older mills did.  Times move on and technology improves.

Svalbonas does pull back from this comparison when he states later that both shell and trunnion-mounted mills can be engineered to have the same level of risk. This has been proven recently. The largest shell-mounted ball mills in the world are designed to stress ranges that are less than the allowable stress ranges for internal welds used by both of the US trunnion-mounted mill manufacturers.

What if the mills are segmented axially to facilitate transportation?  We have undertaken detailed finite element analyses of axially split shell and trunnion-mounted mills.  The increase in stress range at the T joint welds in shell-mounted mills is the same as the increase in stress range at internal welds caused by the segmentation in trunnion-mounted mills.

It can be concluded that shell-mounted and trunnion-mounted mills can be designed with the same safety factors for internal welds for both segmented and unsegmented mills.

5 – Structural Effects of Segmentation

Segmentation of shell-mounted mills is discussed by Svalbonas in [1, 2, 6]. Some options available for segmentation are shown in Figure 1.

Fig. 1. Options for segmentation of riding rings – reproduced from [1] – SAG 2001 Conference, Vancouver

The world’s largest manufacturer of shell-mounted mills, Polysius, uses configurations in Figure 1(a) and (b) with the additional option of site welding to eliminate the circumferential bolted flanges at the can-to-can joint that are shown in Figure 1 (a). There is no structural disadvantage in the design of a shell-mounted mill of the type shown in Figure 1(a) when compared with a similar sized trunnion-mounted mill. In fact, the use of site welding of the shell sections favoured by Polysius is advantageous as bolt failures in flanged mills can lead to structural failures such as those that have occurred recently in South America.

The unclamped length ‘a’ in Figure 1(b) is considered in [1] to require “some extreme designing [that] has not been demonstrated so far” to ensure trouble-free operation that prevents contamination of the bearing. The Ernest Henry SAG mill has this design and although it has suffered contamination of the bearing oil by mill slurry, it has operated continuously since 1997 without a structural failure. The sealing of this region has been improved by a number of pressure grooves and barriers being added to the joint in newer mills. Initial feedback from Hidden Valley (Newcrest), which also has a segmented shell-supported SAG mill, indicates that there are no operational or maintenance issues related to the riding ring (journal) split.

6 – Site welding

Trunnion-mounted mill vendors in their tenders and presentations state that bolting sections together is preferable to site welding because full stress-relieving cannot be completed in the field and the presence of localised residual stresses after site welding require that the shell-mounted mill must be designed using heavier sections. This description of the effects of site welding is an exaggeration. In [6], Svalbonas writes that “site welding of components … is acceptable as long as it is realised that this is not just an alternative erection procedure but it is actually a design change”. He goes on to raise residual stresses but when seen in the context of the quoted sentence, it is seen that there is no fundamental problem with site welding.

The ball mills at Cerro Verde rely on site welding. The calculated stress ranges for a design load of 45% total charge with 40% ball charge were no greater than 30 MPa at the location of the site welds.  The stress ranges conform to all established design codes for weld fatigue such as BS7608.  Both Svalbonas [5, 7] and Farnell-Thompson/FLS [8] consider BS7608 to be too conservative (stringent) for trunnion-mounted mill design as it requires a reduction of the allowable stress range as a function of plate thickness, so site welded joints are not a great risk if designed properly.

Site welding is a viable alternative to bolted joints.

7 – Validation of shell-mounted mill designs: Stress relief and Strain Gauging

We are aware that mill design consultants claim that post-weld heat treatment of welded joints in shell-mounted mills is not as effective as it is in trunnion-mounted mills. This statement is incorrect. Residual stress measurements performed on the Pueblo Viejo project using strain gauges mounted on and adjacent to the T joint welds.  These measurements showed that that the level of stress relief achieved at a the T-joint after post-weld-heat-treatment is as much as would be expected in the end flanges of trunnion-mounted mills.

Strain gauging of shell-mounted mills has been undertaken in the past.  Some results have been published in SAG mill conferences.  Strain gauges were also used to check the design of the CMDIC shell-mounted ball mills. The statement in [1] that shell mounted mills have not been thoroughly tested is incorrect.

CONCLUSION

Shell-mounted mills can be designed to same level of safety (risk) as trunnion-mounted mills. Shell-mounted mills with end walls directly over the main bearings are stiffer than trunnion-mounted mills both globally and locally (at the bearing journals) but this is not a parameter that should influence a mill purchase as both mill types are sufficiently stiff. The segmentation of shell-mounted mills can be designed so that the adjacent welds are not over-stressed.  The axial split in the riding ring (journal) that is not clamped, Figure 1(b), presents a potential design issue that requires attention to detail by the mill designer but we are not aware of any failures emanating from this region. The sealing of this region has been improved substantially over the past 15 years.  Site welding should not be considered as a poor design choice.  The choice of site welding or bolted flanges should be based on a number of factors such as previous vendor experience, expert personnel availability, construction schedule and planning, future maintenance advantages and disadvantages, and stress range rather than the irrelevant fact these these welds will not stress relieved.

Trunnion-mounted mills have been shown to be reliable provided detailing is properly considered. Thickness of flanges, the use of contour flanges versus plate flanges, bolting design and segmentation should all be considered during a tender review. This is particularly the case when North American vendors and design consultants are chosen as they do not model three-dimensional details such as head and shell-axial flanges.

Trunnion-mounted mills are not inherently “better” than shell-mounted mills and vice versa.

REFERENCES

[1] Svalbonas, V., 2001. Difficulties in mill comparisons – Shell supported vs trunnion.In: Barratt et al. (Eds.), Proceedings International Autogenous and Semiautogenous Grinding Technology 2001.

[2] Svalbonas, V., 2002. The design of grinding mills, in Mineral Processing Plant Design, Practice and Control: Proceedings Vol 1: ed. A. L. Mular, D. N. Ialbe, D. J. Barratt, SME

[3] Knecht, J. ; Tew, A. 1999. World’s Largest Shell Supported SAG Mill At The Ernest Henry Concentrator-Operating Experiences And Results. SME Annual Meeting, Denver, Colorado.

[4] Leonard, J., 1999. Evolution to revolution. (third generation shell supported mining mills). Mining Magazine, April.

[5] Svalbonas, V., Berney-Fiklin, J., 2006. History of weld design for grinding mills (a participant’s view). In: Barratt et al. (Eds.), Proceedings International Autogenous and Semiautogenous Grinding Technology 2006.

[6] Svalbonas, V., 1999, Mechanical design of large grinding mills – AG/SAG Part 1, prepared for Mineral processing and hydrometallurgical plant design – World’s best practice, Australia Mineral Foundation.

[7] Svalbonas, V., Fresko, M., 2002. How safe are your recent mills?: The compatibility between FEA and welding codes. SME Annual Meeting, Phoenix, Arizona.

[8] Farnell-Thompson Applied Technologies & FLS:  From tender negotiations and design reviews on several projects.

The title of the paper is similar to the title of the 1970s book “Zen and the Art of Motorcycle Maintenance”, which is itself a play on the earlier book “Zen and the Art of Archery”.  A whole series of similarly titled books have been released.  The motorcycle maintenance book presents the philosophical concept of Quality that cannot be described in words but is something that must be felt or experienced. It is the parent entity to subjective and objective thought; it is metaphysical unlike the concept of quality we use in manufacturing.  Quality in is the most base potential in the universe that pushes everything towards higher levels of Quality.  This is a circular definition that makes it a touchy-feely concept that cannot readily be applied to engineering and is even less comprehensible than quantum mechanics in explaining the universe we live in.  Zen in the book appears to be about living in the moment, i.e., getting a spiritual experience out of something as mundane as motorcycle maintenance.  All this is very good in a fictional book but what does Zen have to do with specification writing?  There may be some engineers who undergo a spiritual experience when writing a mill specification but they would be rare.  Zen in this paper is more likely to indicate the need for an over-arching philosophical framework in specification writing, e.g., prescriptive vs performance specifications, risk mitigation, have the goal in mind, etc., and if interpreted that way, the paper appears to offer some promise of assisting the mill specification writer.  A better understanding of what is important in a mill specification can only be a good thing but does the paper provide value and is it consistent with the owner’s or “end user’s” goals, which are key assessment criteria used in the paper to assess specifications?

Mills are a focal point in a plant because they are usually on the critical path of the construction schedule, they represent a single major capital purchase, their availability determines the throughput of the plant and their performance is critical to a project’s financial success.  They are not a commodity.  They are tailor-made, steel lined, rotating structures for grinding ore.  They can weigh more than 2000 tonnes and consume of the order of 30 MW of power continuously.  Not only are the mills critical to the success of a process plant but their drives are too.  If either a large grinding mill or the ring motor that drives it fails, the costs to the owner in terms of lost production can amount to hundreds of millions of dollars.  Grinding mill designs are not all equal – trunnion-mounted mills are very different to shell-mounted mills in both construction and performance.  Unfortunately, modern mills do fail despite vendor claims to the contrary.  Since the late 2000s there have been partial and catastrophic failures in mills designed using modern technical “industry standards” in Australia, Argentina, Chile and Peru.  So what should be the owner’s goals be when purchasing mills?  The owner and its agents should be focussed on procuring mills that will (a) grind the required quantity of ore to the size that the process requires and (b) have a probability of failure that is sufficiently low so that it balances or offsets the consequence that a major failure would have on the owner’s business.  So, very large mining companies may view a purchase differently from a small or mid-tier miner because the consequences of a failure to each business are different.  The owner’s goal reduced to its essence is to manage risk and reward over the life of the project.  With this owner’s goal defined, let us assess some points in the Zen of specification writing paper.

1 – Writing Style

The paper states that the style of writing is important in developing a good mill specification.  It notes that many specifications are amalgamations of many other specifications and that this may result in a disjointed document that has conflicting requirements.  It highlights that the bespoke aspect of mills can lead to an overly detailed specification that could be “inscrutable to the reader”.  Nemchek observes that the mills are not built by the vendor in their own factories but are sub-contracted out to sub-suppliers and that the lack of direct feedback from the shop floor to the vendor can result mill specifications that are not consistent with modern manufacturing practices but we don’t really understand this logic.  Interestingly, Nemchek states that a typical mill specification is approximately 500 pages long!

It is clear that a well constructed specification document will make it easier for everyone involved to manage the mill contract but the paper’s approach of simplicity, brevity and clarity in writing are not sufficient to achieve a good specification.  The paper does not consider who will write the specification.  A good specification requires that writer understands what they are writing and why they are writing it.  Thus, an investment is required on the part of the owner or their engineering contractor to train or hire experts who know (a) the capabilities of all the mill vendors in detail; (b) what are the likely pitfalls to be faced with each vendor and each mill type; and (c) how to address these issues in a document.  Furthermore, the mill specification is rarely passed on to vendors’ sub-suppliers as stated in the paper.  Mill vendors have their own fabrication specifications that they send to their sub-suppliers.  What the specification must ensure is that owner requirements that are additional to the requirements in the mill vendor’s standard fabrication specifications are included in the vendor’s specifications as addenda and that these are highlighted and explained to the sub-supplier prior to the commencement of manufacturing.

We have never seen a 500 page mill specification but that does not mean they do not exist.  The typical, detailed mill technical specifications we have worked with or written are between 50 and 70 pages long including the data sheets.  Even the over-the-top specification for the Sept-Iles Expansion was of the about 110 pages, not 500, and this document is quite easy to read and understand.  If 500 page technical specifications exist, then they must be rejected by the vendors and the engineering company that wrote these should not be paid for the redrafting.  What we have found that is more concerning are 13 page specifications without clause numbering written without any consideration of risk faced by the owner that were developed by the mini-Fluor/Bechtel pretenders that rose to prominence during the mining boom.  These are performance specifications that take brevity to a new level.  They are a product of EPCM contractors taking a detailed specification and removing anything that could result in any risk to themselves.  It was enormously difficult manage these specifications as locating a particular clause involved counting lines on a page rather than referencing clause numbers.  You can imagine the frustration this caused when discussing major technical issues by phone or by email with owners, vendors and sub-suppliers on different continents.

A further major issue arises in how mill specifications are used by the various parties involved.  In particular, the role of commercial managers who have no practical technical knowledge can completely nullify the technical specification with a stroke of a pen.  Their need to justify their role by screwing down the vendor’s price is amusing when we have been told by several vendors that they know who these people are and so they increase their initial prices accordingly to provide the win required to clinch the contract.  If this is too hard to swallow, consider how many mill vendors lost money or didn’t make super-profits during the mining boom and how many commercial managers were bragging about what a great “discounts” they obtained.  Often the discount comes at a price such as relaxed acceptance criteria and the engineer responsible for the mills is left to pick up the pieces whilst the commercial manager dines with the mill vendor, sometimes on the owner’s tab.

Writing style is important but not as important as content or the negotiations that precede the letting of the contract.  Writing in active rather than passive voice is not going to matter much if the specification neglects to include the requirement for special tools, vibration monitoring systems, lifting lugs, hold-down bolts, etc.

2 – Mutual Benefit

The paper states that “a good business deal makes everyone happy” and that a “good specification” is mutually beneficial to the owner, the engineering firm and the vendor.  In practice this means that the owner will want a detailed specification that minimises the risk to an appropriate level over the life of the plant.  This will include stress levels, mill type, performance, machining requirements, quality acceptance criteria, etc.  The engineering company will want to avoid any risk, so they will want a performance specification with little or no technical detail as in the DC 3 example in the paper.  The vendor will want no interference by the owner in their processes at all and thus also prefer a minimal specification with justifications such as “trust us; we have done it all before; our mills don’t fail so what are you worried about?”.  Prior to 2009, the mantras describing prescriptive specifications as unnecessary from the engineering companies and vendors were hard to refute as there were no “modern mill” failures.  Since 2009, there have been a number of failures in large trunnion and shell-mounted mills.  Leaving the design and manufacturing details entirely up to the vendor is not a legitimate option for a diligent owner who is obliged to look after their shareholders’ interests when managing risk.  Thus a specification is unlikely to be “good” for all of the entities involved in a mill purchase.

The paper states that specifications presume an adversarial relationship will exist between the owner and the vendor and that the vendor will “always provide the lowest quality at the highest price”.  Well, why would a vendor do anything else?  That is how they will maximise their profit, which is their primary goal.  This should not be a contentious statement as profit maximisation underpins the whole market economy.  The paper is wrong in assuming that goals  of all the participants in a mill contract will be aligned.  Even worse is that the paper does not consider relative risk.  When a mill fails prematurely say 5 or 10 years after commissioning, the engineering contractor will not face any consequence and the vendor will likely be paid to provide replacement components.  It is only the owner who is likely to face losses and these can be catastrophic for their business.  Furthermore, an adversarial relationship between owner and vendor will usually arise when something goes wrong in the manufacturing or delivery.  Even then, it is when the manufacturing fault is foreseeable that friction between the parties will occur.  We have not seen any technical specification that presumes such an adversarial relationship apriori.

Nemchek states that “mill specifications tend to treat quality requirements as an ointment that, if applied insufficient quantity, will soothe clients.  Often, these requirements are written in such a way that the client may add any testing desired, at the supplier’s expense, with acceptance criteria based on the client’s satisfaction”.  We have not seen any specifications of this type.  The paper goes on to say “[q]uality professionals generally acknowledge that quality cannot be improved by inspecting after the fact. In other words, an out-of-control casting, fabrication or machining process will yield the same results, regardless of how often or how closely it is inspected cannot be improved”.  This rings true but it is not meaningful.  Consider the casting example below.  The casting was presented by the vendor for final inspection and acceptance by the owner.  The flaw in the flange would no doubt have resulted in a premature failure had the owner not conducted a detailed inspection to (a) find the flaw and (b) insist that the criticality of the flaw be investigated by grinding.  The repair, whilst unsightly, was far more preferable than a broken head.

3 – Who pays for Quality?

Nemchek states that some specifications have open ended quality criteria and these can result in testing being added at the vendor’s expense.  Whilst this may occur, we have not seen it.  Our experience is that owners, their agents and the vendors all have the opportunity to come to an agreed specification prior to an offer being made final by the vendor and accepted by the owner.  This includes all the testing required to assess mill components.  Mill vendors are not babes in the woods when it comes to understanding their exposure.  However, once a component is found to be outside the agreed specification acceptance criteria, shouldn’t the owner have the right to insist on additional testing and/or compensation before accepting the component?  It seems only fair.  What often happens during mill manufacturing is that vendors shift the goal posts by claiming that the faulty component is “fit-for-purpose” and so do not reject the component.  This becomes an issue that cannot always be resolved technically.  The solution is to have a project manager with a backbone and a contract that gives them the power to influence the process when things go wrong  and to direct changes if necessary.

It should be noted that fitness-for-purpose means different things to different people.  The cars below both provide transportation for four adults.  If the number of adults were the only criterion, then both vehicles would be fit-for-purpose but which one would you buy?

(Image from “The Sculptor” commercial by Peugeot)

4 – Seeking the Supplier’s Input

This is a well written section.  The mill vendor’s buy-in to the process is critical to the success of the project.  However, it is important to be aware that a mill vendor will not be and possibly cannot be as open in their tender as the owner requires.  Mill vendors often provide a list of potential sub-suppliers for a component and include the rider “or equal”.  There are not many foundries that are “equal” to Ferry Capitain, Siempelkamp or Sidenor.  Similarly, all fabricators are not equal.  The owner’s team must ensure that the sub-suppliers chosen by the vendor are capable of delivering components that meet the owner’s requirements.  This is often neglected in tender negotiations.

CONCLUSION: Is “Zen and the Art of Specification Writing” Valuable?

The paper presents an approach to improving specifications for grinding mills.  Its focus on a holistic approach to specification writing is worthy but the paper does not consider the mill procurement process in sufficient detail in reaching its conclusions.  A substantial part of the paper is devoted to grammatical issues such as active and passive voice, simplicity, brevity and clarity.  It includes a “Miranda warning” twice in the paper, i.e., “anything the specification writer says can and will be used against them”.  This is inconsistent with the paper’s criticism that current mill specifications presume the relationship between the owner and the vendor will be adversarial.  The paper discusses too little or too much detail being included in the specification but it does not indicate what details are important.  It does not consider the relative risks that the parties involved in the process face.  Its take on quality is quite naive as shown in the casting photos in this post.  Like the novel, Zen and the Art of Motorcycle Maintenance, the paper is an attempt to transcend the classical approach (philosophy) of specification writing (motorcycle maintenance) rather than providing practical examples of what is needed and any benefit the reader gains from the paper (book) depends on the reader’s predisposition to self-analysis.  As a tool for improvement of all mill specifications, the value of the paper as it was published in 2011 is likely to be similar to the limited effect that the “Zen and the Art of . . . ” books have on improving the overall psychological well-being of western society.

Possibly the greatest omission in the paper is the lack of comment on the Scope of Work.  This is the most important part of the specification.  If the scope is wrong then any further effort in writing the specification or fostering good relationships with the vendors or testing the mill components will be wasted.  This could be included in the paper philosophically as “putting first things first”.

Our contribution to the “Zen” of specification writing would be this.  If you are starting a new specification, then some of the most relevant and useful spiritual experiences you can have are found during commissioning of mills out in the desert in summer or repairing mills late in the night in freezing conditions.  These experiences will increase your desire to learn as much as you can on the technical, commercial and managerial aspects of the mill procurement process.  They will improve your mill specification immensely.