A vertical roller mill is a key machine in cement and mineral powder production. Among its many components, the grinding table, usually called the disc body, is one of the most critical parts because it directly carries the grinding load and transmits it to the speed reducer and the foundation. The disc body must operate within a severe environment that includes high temperatures, dusty gas, high compressive loads, dynamic impacts, and cyclic start-stop conditions. Any structural failure of the disc body can stop production immediately and may create a serious safety risk. I participated in the technical investigation of a disc body failure that occurred in a plant where the mill had been put into service in May 2020 and where abnormal vibration and repeated mill trips appeared in July 2023. After shutdown, inspection revealed cracks in the disc body. This article describes how I analysed the failure, how I reached the root cause, and how the disc body was repaired and returned to safe temporary service.
Observed Cracking Condition
During the shutdown inspection, I found two cracks on the outer large face of the disc body. The two cracks were distributed approximately 90° apart along the circumference. The crack initiation region was located at the outer diameter, near the lifting-lug transition, and it extended from the outer lugs upward to the upper machined face. The measured crack length was about 600 mm and the crack opening was about 2 mm. Ultrasonic inspection indicated that the crack depth had not yet penetrated completely through the wall to the internal cavity of the disc body. After removal of the wear liners, I inspected the upper machined surface and found two more cracks in the same radial planes as the outer cracks. These upper-face cracks extended approximately 350 mm from the maximum outer diameter toward the centre of the disc body. The crack morphology suggested a combination of thermal fatigue and mechanical crack growth with multiple initiation sites.
The following illustration shows a typical example of a ductile iron casting similar to the family of components discussed in this failure analysis.
Failure Analysis Strategy
Because a sudden failure of this type creates serious concerns for the owner, it was essential to identify the root cause quickly and with a high level of confidence. I adopted a comprehensive method that combined three main investigation directions: casting quality, mechanical strength, and operation and maintenance. The table below summarises the evaluation matrix.
| Evaluation domain | Evidence reviewed | Conclusion |
|---|---|---|
| Casting quality | Material certificate, retained samples, chemical analysis, tensile testing, metallographic testing, ultrasonic inspection records | No nonconformity detected |
| Mechanical strength | Finite element analysis under design load, hydraulic pressure records, actual load calculation, safety factor calculation | Mechanical strength was adequate |
| Operation and maintenance | Control room logs, inlet gas temperature, feed moisture content, starting and stopping frequency, cooling method | Thermal overloading was the primary cause |
This structured approach is especially important for large ductile iron castings, because a hasty conclusion can lead either to an unnecessary replacement or to an unsafe repair. The analysis had to answer a simple question: was the crack caused by bad material, bad design, or bad operation? I discuss each domain below.
Casting Quality Verification
The disc body was manufactured from EN-GJS-400-18 ductile iron according to DIN EN 1563. This material belongs to the family of ferritic ductile iron castings and is selected for thick-section castings that require high ductility, impact resistance, and machinability. For ductile iron castings used in grinding equipment, the microstructure is normally required to be predominantly ferrite with spheroidal graphite. The specification used in this project also prescribed a minimum nodularity of 90%, which corresponds to a type 2 graphite classification.
I first reviewed the original casting documentation using the disc body serial number. The recorded composition, mechanical properties, metallographic results, and ultrasonic inspection reports all matched the design requirements. To avoid relying only on historical records, I also located the retained separately-cast test samples from the same production batch and arranged for independent tests. This step is important because retained samples allow direct confirmation that the actual cast metal delivered to the site has the required properties.
Chemical Composition Requirements and Measured Values
The table below gives the chemical composition requirements for the disc body material. The main controls are carbon, silicon, manganese, phosphorus, and sulphur.
| Element | Required range / maximum in wt-% |
|---|---|
| C | 3.40 – 3.65 |
| Si | 2.20 – 2.80 |
| Mn | 0.20 – 0.35 |
| P | ≤ 0.035 |
| S | ≤ 0.015 |
| Residual elements | Controlled according to the foundry specification |
The measured composition from the retained sample is shown below.
| Element | Measured value in wt-% |
|---|---|
| C | 3.54 |
| Si | 2.51 |
| Mn | 0.30 |
| P | 0.013 |
| S | 0.004 |
I also calculated the carbon equivalent using the measured values. A common expression for the carbon equivalent of ductile iron is:
$$ CE = w(\mathrm{C}) + \frac{1}{3}\left[ w(\mathrm{Si}) + w(\mathrm{P}) \right] $$
Substituting the measured values gives:
$$ CE = 3.54 + \frac{2.51 + 0.013}{3} \approx 4.38\% $$
This carbon equivalent is normal for the EN-GJS-400-18 grade. It indicates a stable eutectic solidification structure and is consistent with the production of well-nodularised ductile iron castings.
Mechanical Property Requirements and Measured Values
The required mechanical properties of EN-GJS-400-18 are given below. The disc body was specified with a minimum yield strength of 240 MPa, a minimum tensile strength of 370 MPa, and a minimum elongation of 12%. Hardness was also controlled to ensure consistency with the grinding application.
| Property | Required value |
|---|---|
| Yield strength \(R_e\) | ≥ 240 MPa |
| Tensile strength \(R_m\) | ≥ 370 MPa |
| Elongation after fracture \(A\) | ≥ 12% |
| Brinell hardness | 120 – 175 HBW |
| Nodularity | ≥ 90% (type 2) |
| Graphite size | Type 5 – 6 |
| Pearlite content | < 10% |
The retained sample was tested by an independent laboratory. The tensile results are summarised in the table below. Three test specimens were used.
| Tensile parameter | Specimen 1 | Specimen 2 | Specimen 3 | Requirement |
|---|---|---|---|---|
| Tensile strength / MPa | 402 | 402 | 402 | ≥ 370 |
| Yield strength / MPa | 267 | 268 | 268 | ≥ 240 |
| Elongation / % | 19.5 | 18.5 | 19.0 | ≥ 12 |
| Reduction of area / % | 18 | 18 | 19 | Not specified |
All measured values exceed the minimum requirements. The yield strength was about 11% above the minimum, the tensile strength was about 9% above the minimum, and the elongation was well above the specified value. This is what I would expect from well-produced ductile iron castings with a fully ferritic matrix and high nodularity.
Metallographic Examination
The metallographic examination was performed on a sample taken from the retained test block. The results below confirm that the graphite shape, nodularity, and matrix structure were all acceptable for this grade.
| Metallographic feature | Requirement | Measured result |
|---|---|---|
| Graphite shape | Spheroidal type VI | Type VI |
| Nodularity | ≥ 90% | 93% |
| Graphite size | 90% type 6 and 10% type 5 | 64% type 6, 30% type 7, 6% type 8 |
| Pearlite content | < 10% | Less than 10% |
For ductile iron castings, nodularity is one of the most important parameters because poor nodularity reduces ductility and crack resistance. The measured nodularity of 93% is excellent. The graphite size is also in the expected range for a large section casting.
Ultrasonic Inspection Records
The original manufacturing procedure required 100% ultrasonic inspection of the upper plane of the disc body before shipment. The acceptance level was Level 2 according to EN 12680-3:2011. I reviewed the original ultrasonic report, and there were no recorded rejectable indications. This makes an internal shrinkage defect or major casting discontinuity very unlikely as the root cause of the crack.
After reviewing all casting data, I found no evidence of any manufacturing nonconformity. The material, microstructure, and cast-in quality of the ductile iron castings were acceptable and met the drawing requirements.
Mechanical Strength Verification
After confirming that the material was acceptable, I turned to the mechanical design. The disc body is extremely large and heavily loaded, so a purely mechanical overload can be an obvious suspect. I performed a finite element analysis using a single-roller maximum projection pressure of 1200 kN/m². The model included the full three-dimensional geometry of the disc body, the contact area of the grinding track, the support boundary at the hub, and the hydraulic loading system. The analysis assumed linear-elastic behaviour with the room-temperature material properties specified for EN-GJS-400-18.
The finite element results showed that the maximum von Mises stress in the disc body under the full design load was only about 33 MPa. The location of the maximum stress was in the transition zone near the outer diameter and the lifting lugs. This location coincided with the observed crack position, but the stress magnitude was very low compared with the material yield strength. The stress concentration factor at that transition is significant, but the nominal stress level is low enough that the design is safe under normal room-temperature conditions.
The ratio between the yield strength and the maximum applied stress gives the safety factor. For the design condition:
$$ n_{\mathrm{design}} = \frac{R_e}{\sigma_{\mathrm{design}}} = \frac{240}{33} \approx 7.3 $$
A safety factor of more than seven is normally considered very conservative for a static mechanical design. Therefore, the disc body was not under-designed for mechanical load at room temperature.
I also checked the actual operating load using the mill control system records. The hydraulic pressure in the upper chamber was about 10 MPa, while the pressure in the lower chamber was about 2.3 MPa. From these pressures I calculated the actual grinding roller projection pressure at the time of operation:
$$ p_{\mathrm{actual}} \approx 920\ \mathrm{kN/m^2} $$
Because the actual roller pressure was lower than the design value, the maximum stress in the disc body under actual mechanical load was:
$$ \sigma_{\mathrm{actual}} = 33\ \mathrm{MPa} \times \frac{920}{1200} \approx 25.3\ \mathrm{MPa} $$
The corresponding actual safety factor was therefore:
$$ n_{\mathrm{actual}} = \frac{R_e}{\sigma_{\mathrm{actual}}} = \frac{240}{25.3} \approx 9.5 $$
| Condition | Projection pressure / kN·m⁻² | Maximum von Mises stress / MPa | Safety factor |
|---|---|---|---|
| Design condition | 1200 | 33 | 7.3 |
| Actual operation | 920 | 25.3 | 9.5 |
These results demonstrate that the disc body had ample static strength for the mechanical loads acting on it. The failure could not be explained by ordinary mechanical overloading.
Operational and Thermal Assessment
Because the casting quality and mechanical strength were both acceptable, I focused the investigation on the operating environment. I examined the plant control room records and the maintenance log for the period immediately before the crack was detected. The data revealed several important facts.
First, the moisture content of the feed material was often above 20%. High moisture forces the hot gas system to work harder to dry the material inside the mill, and it tends to raise the required gas inlet temperature. Second, during July, the inlet gas temperature was generally above 600°C. The highest recorded temperature was more than 750°C. Third, the plant ran in an off-peak mode that caused the mill to start and stop several times every day. Fourth, after shutdown, the operators opened the air door and supplied compressed air directly into the mill to cool the mill quickly. The compressed air was aimed at the disc body, causing strong local cooling of the hot upper surface.
This combination of very high inlet temperature and rapid artificial cooling was dangerous for a material such as EN-GJS-400-18. Published studies on the high-temperature behaviour of ductile iron castings show that the yield strength and tensile strength decrease slowly as temperature rises from room temperature to about 400°C. Above 400°C, the strength falls much more rapidly. At 500°C, the strength is roughly 50% of its room-temperature value. At 600°C, it is about 30%. At 700°C, the retained strength may be only about 10% to 15% of its original value. The table below summarises this behaviour.
| Temperature / °C | Approximate retained yield and tensile strength / % | Approximate yield strength for EN-GJS-400-18 / MPa | Safety factor at actual stress 25.3 MPa |
|---|---|---|---|
| 20 | 100 | 240 | 9.5 |
| 400 | ~90 | ~216 | ~8.5 |
| 500 | ~50 | ~120 | ~4.7 |
| 600 | ~30 | ~72 | ~2.8 |
| 700 | ~10 – 15 | ~24 – 36 | ~1.0 – 1.4 |
This table is especially important for understanding the root cause. The finite element calculation showed that the mechanical stress in the disc body was only about 25 MPa under normal operation. At first, this seems safe. But if the metal temperature rises to 700°C, the yield strength can fall to a value near or even below 25 MPa. The safety factor then becomes close to unity, or even less than unity. A small additional stress from thermal gradients or phase transformation can push the combined stress above the available yield strength and initiate a crack.
The thermal stress caused by rapid cooling can be estimated with a simple restrained-body expression:
$$ \sigma_{\mathrm{th}} = \frac{E \alpha}{1 – \nu} \Delta T $$
In this expression, \(E\) is the elastic modulus, \(\alpha\) is the coefficient of thermal expansion, \(\nu\) is Poisson’s ratio, and \(\Delta T\) is the temperature difference between the hot interior and the cooled surface. For ductile iron, the elastic modulus is roughly 169 GPa and the coefficient of thermal expansion is about \(12 \times 10^{-6}\ \mathrm{K^{-1}}\). If the surface is suddenly cooled by compressed air while the interior remains hot, a temperature difference of only 50 K can produce an elastic thermal stress of:
$$ \sigma_{\mathrm{th}} \approx \frac{169 \times 10^9 \times 12 \times 10^{-6} \times 50}{1 – 0.275} \approx 140\ \mathrm{MPa} $$
This simple estimate is conservative because it assumes fully restrained expansion, but it shows the order of magnitude of thermal stress that can be generated by rapid cooling. Even if actual thermal stress is only a fraction of this idealised value, it is still large compared with the 25 MPa mechanical stress. In ductile iron castings, such thermal stresses are often the controlling load case during start-up, shutdown, and emergency cooling.
The disc body material is primarily ferritic ductile iron. According to the iron-carbon phase diagram, ferrite begins to transform to austenite at approximately 720°C. This transformation is not only a metallurgical change; it is also accompanied by a volume change and by carbon diffusion between the graphite nodules and the matrix. In a large casting, the outer layer may transform while the inner core is still cooler and remains ferritic. The transformation strain creates an additional internal stress that is superimposed on the mechanical and thermal stresses. The phase transformation can be expressed simply as:
$$ \mathrm{Ferrite} \quad (\mathrm{BCC}) \xrightarrow{\sim 720^\circ\mathrm{C}} \mathrm{Austenite} \quad (\mathrm{FCC}) $$
For ductile iron castings, the normal stress-relief annealing range is 550°C to 600°C. The recommended practice is to hold the component in that range for 2 to 8 hours and then cool it slowly at a cooling rate below 30°C per hour. If the component is repeatedly exposed to temperatures higher than 550°C and then cooled rapidly, the beneficial stress-relieving effect is lost. In addition, very high temperatures can cause graphitisation and coarsening of the graphite structure, which further reduces the strength and hardness of the material.
The table below summarises the thermal parameters that I used in assessing the disc body behaviour.
| Parameter | Value / range |
|---|---|
| Stress-relief temperature for ductile iron castings | 550 – 600 °C |
| Holding time | 2 – 8 h |
| Recommended slow cooling rate | < 30 °C/h |
| Temperature of ferrite-to-austenite transformation | ~720 °C |
| Maximum safe continuous operating temperature for EN-GJS-400-18 | < 400 °C |
The operating records show that the mill was frequently run with inlet gas temperatures above 600°C and sometimes above 750°C. The disc body surface in direct contact with the hot gas must have reached temperatures near the phase transformation range. At the same time, the interior sections and the area around the water-cooled or air-cooled support could be at a much lower temperature. The resulting temperature gradient was severe.
The frequent start-stop cycles also contributed to the damage. During normal operation, the disc body is heated slowly. During shutdown, the operator supplied cold compressed air directly onto the hot disc body, which caused rapid cooling of the surface. This produced cyclic thermal fatigue. In a large rotating disc, the outer lugs and the upper-face transition have a higher stress concentration than the rest of the body. Micro-cracks initiated at these locations and then propagated as a result of repeated thermal and mechanical cycling.
The crack growth can be described by the Paris law for fatigue propagation:
$$ \frac{da}{dN} = C \left( \Delta K \right)^m $$
where \(a\) is the crack length, \(N\) is the number of cycles, \(\Delta K\) is the stress-intensity-factor range at the crack tip, and \(C\) and \(m\) are material constants. The thermal cycling caused \(\Delta K\) to exceed the threshold value, and the crack propagated progressively from the outer surface toward the centre. The final observed crack length represented the cumulative damage after many thermal and mechanical cycles, rather than a single catastrophic overload.
Root Cause Determination
By combining the evidence from the casting review, the finite element analysis, and the operation records, I reached the following root cause conclusion. The disc body was correctly manufactured from high-quality ductile iron castings and had adequate mechanical strength for normal room-temperature design conditions. The failure was caused by an operating condition that exceeded the thermal capability of the material. The inlet gas temperature above 600°C and occasional peaks above 750°C reduced the yield strength of the material dramatically. At the same time, rapid cooling with compressed air generated large thermal gradients and cyclic thermal fatigue. These two factors together reduced the local load-bearing capacity of the ductile iron below the combined mechanical and thermal stress, causing cracks to initiate in a region of geometric stress concentration and then propagate inward.
| Contributory factor | Role in failure |
|---|---|
| Inlet gas temperature above 600°C | Reduced yield and tensile strength; primary cause |
| Occasional inlet gas temperature above 750°C | Approached phase transformation temperature; caused severe strength loss |
| Direct compressed-air cooling after shutdown | Created large thermal gradients and thermal fatigue |
| Frequent daily start-stop operation | Increased the number of thermal cycles and contributed to crack growth |
| Outer lug geometry transition | Localised the crack initiation site |
| Casting quality, material grade, and design strength | No deficiency found |
This root cause analysis gave me confidence that the crack was not caused by a defective casting. It also showed that a properly executed repair could allow the mill to operate safely for a limited period while a replacement disc body was being manufactured, provided that the thermal operating limits were strictly controlled.
Repair and Strengthening Measures
When a large ductile iron component such as a grinding table cracks, the first decision is whether to repair or replace it. Because the crack had not fully penetrated through the wall to the inner cavity, the remaining sound section was still large enough to carry the load for temporary service after a controlled repair. I worked with the maintenance team to develop a repair procedure that would stop crack growth, restore some structural continuity, and allow the mill to run until the new disc body arrived.
Crack Arrest and Reinforcement
The first step was to prevent further propagation of the cracks. I specified holes to be drilled at the tips of each crack. These stop holes act as crack arrestors by increasing the root radius at the crack tip and reducing the stress intensity factor. The reinforcement plan required machining a number of slots perpendicular to the crack direction and inserting steel connection plates into these slots. The plates were then welded to the disc body. This arrangement provides a mechanical bridge across the crack and shares part of the load with the steel plates. The reinforcement plate orientation is critical: it must be perpendicular to the crack path so that the plate supports the opening direction of the crack.
The table below lists the main repair steps.
| Step | Action | Purpose |
|---|---|---|
| 1 | Clean the cracked area, removing oil, dust, and loose scale | Prepare the surface for inspection and welding |
| 2 | Drill stop holes at both ends of each crack | Prevent crack propagation |
| 3 | Machine slots perpendicular to the crack direction | Create space for reinforcement plates |
| 4 | Insert steel reinforcement plates | Restore structural continuity |
| 5 | Preheat the repair zone to 500 – 700 °C | Reduce welding stress and avoid cracking |
| 6 | Weld the plates using EZNiFeCu electrodes | Produce a ductile, machinable weld metal |
| 7 | Control welding current, voltage, and travel speed | Minimise heat input and distortion |
| 8 | Apply post-weld heat treatment | Relieve residual stresses and stabilise the microstructure |
| 9 | Cool slowly below 30 °C/h | Avoid new thermal gradients |
| 10 | Grind the weld flush and perform non-destructive testing | Verify the quality of the repair |
Welding Consumable Selection
Welding of ductile iron castings is challenging because the material is sensitive to rapid heating and cooling. If the welding electrode is not chosen carefully, the weld can form hard, brittle martensite or cracks in the heat-affected zone. For this repair, I selected a nickel-iron-copper electrode classified as EZNiFeCu, commonly known as Z508. This electrode has good crack resistance, produces a weld metal with favourable ductility, and remains machinable after welding. The selection of the correct electrode is one of the most important decisions when welding any ductile iron casting, because the weld metal should ideally have a coefficient of thermal expansion close to that of the cast iron and should be able to tolerate some dilution from the base metal.
| Welding characteristic | Selected value / recommendation |
|---|---|
| Electrode type | EZNiFeCu (Z508) |
| Main alloy constituents | Nickel, iron, copper |
| Advantages | Good crack resistance, good machinability, suitable for ferritic ductile iron castings |
| Preheating temperature | 500 – 700 °C |
| Post-weld treatment | Annealing / stress relief followed by slow cooling |
Preheating and Welding Process Control
Preheating is essential for welded repairs of ductile iron castings. In this case, the preheat temperature range was 500°C to 700°C. The entire weld area and the adjacent base metal must be preheated uniformly. This high preheat reduces the temperature difference between the weld pool and the surrounding metal, lowers the cooling rate, and prevents the formation of hard microstructures. Temperature should be monitored with a contact thermometer or an infrared pyrometer.
During welding, I instructed the welders to use a low current and a short arc. Excessive heat input increases the risk of burning the graphite, producing porosity, or causing distortion. The welding rod should be moved smoothly, and the bead width should be kept moderate to avoid concentrating too much heat in one place. After each pass, the weld should be allowed to cool slowly; never use forced air or water to cool the weld area.
After welding, the component must be post-weld heat treated. I recommended a stress-relief anneal at 550°C to 600°C with a holding time of 2 to 8 hours, followed by slow cooling at a rate below 30°C/h. This treatment reduces residual stresses, stabilises the metallurgical structure, and restores some of the ductility lost during welding. It is particularly important in large ductile iron castings because residual stresses from welding can combine with service stresses and initiate new cracks.
Non-Destructive Testing After Repair
After the repair was completed, the weld area was ground smooth and subjected to non-destructive testing. I specified liquid penetrant testing of all accessible surfaces to detect surface cracks, and ultrasonic examination of the repaired region to check for lack of fusion and internal defects. The acceptance criteria were the same as the original drawing requirements. This testing is necessary to prove that the repair is structurally sound before the mill is returned to service.
Operational Recommendations for Future Service
The repair of a damaged disc body is only a temporary solution. It does not change the fundamental limitation of the material. To prevent another failure, I strongly recommended the following operating rules for the plant.
First, the inlet gas temperature to the mill must be controlled so that the metal temperature of the disc body does not exceed 400°C. Above this temperature, the yield strength of EN-GJS-400-18 begins to fall rapidly. The mill control system should include an interlock that alarms when the inlet gas temperature approaches 400°C and automatically reduces the hot gas temperature or trips the mill if the limit is exceeded.
Second, the plant must avoid rapid cooling of the disc body after shutdown. Compressed air should never be blown directly onto the hot disc body surface. Natural cooling with the mill stopped, or cooling by the normal gas system with the temperature carefully controlled, should be used instead. If fast cooling is required for maintenance reasons, the cooling rate should be limited and the surface temperature monitored.
Third, the mill should be operated in a stable condition. The feed moisture, feed particle size, and feed rate should be kept within the design envelope. The grinding pressure should be set according to the manufacturer’s recommendation, and the mill vibration level should be continuously monitored. Frequent unnecessary start-stop cycles should be avoided because each cycle adds thermal fatigue damage to the disc body.
Fourth, the repaired area should be inspected regularly. I recommended a visual inspection every month and an ultrasonic inspection every three months, at least until the replacement disc body is available. Any sign of crack re-growth should be treated as an emergency and the mill should be stopped.
Finally, the plant should order a new spare disc body immediately. The replacement material should be the same ductile iron grade, but the design should be reviewed to improve the transition radius at the outer lifting lugs in order to reduce the local stress concentration. If the process will continue to use very high inlet temperatures, the designer should consider an alternative material with better high-temperature strength, or provide additional insulation or cooling features in the disc body. However, for standard ductile iron castings, the simplest and most reliable safeguard is to respect the maximum service temperature.
Discussion of Thermal Fatigue in Ductile Iron Castings
This failure is a good example of thermal fatigue in ductile iron castings. Thermal fatigue occurs when a material is subjected to repeated heating and cooling cycles. It is not controlled by a single material property. Instead, it is a combination of high-temperature yield strength, ductility, toughness, thermal expansion, thermal conductivity, and resistance to oxidation and metallurgical change.
For ductile iron castings, the spheroidal graphite structure provides better ductility and thermal fatigue resistance than flake graphite iron. However, the ferritic matrix still has a limited high-temperature capability. When the operating temperature rises above 500°C, the oxide scale formation increases, the matrix strength decreases, and the graphite-matrix interface becomes a preferred path for crack initiation. Under cyclic thermal stress, small cracks form at the surface and then propagate through the casting.
The presence of graphite nodules in ductile iron castings also influences crack growth. Graphite nodules act as crack arrestors because they blunt the crack tip and reduce the local stress intensity. This is one reason why high nodularity is so important. The disc body in this case had a nodularity of 93%, which means that the graphite was well formed and had a good ability to resist crack propagation. If the nodularity had been poor, the crack would likely have grown much faster.
In this incident, the crack started at the outer lifting lug region, where the geometry is irregular and the stress concentration is highest. The crack then propagated along the radial direction on the upper face and along the axial direction on the outer face. The crack path was consistent with the direction of the maximum thermal stress. The outer surface cooled faster than the interior, producing tensile stresses at the surface. These tensile stresses opened the cracks and drove them inward.
The crack length of approximately 600 mm on the outer face and 350 mm on the upper face indicates that a significant portion of the disc body section had been weakened. The fact that the crack had not penetrated into the internal cavity was fortunate. It was still necessary to calculate whether the remaining sound ligament could support the grinding load during temporary operation. Because the mechanical stress was only 25 MPa even under full load, the remaining section was sufficient, provided that the thermal limit was respected.
The Importance of Retained Samples and Material Traceability
One of the most valuable aspects of this investigation was the existence of retained samples and complete material traceability. Without the original chemical analysis, tensile test data, metallographic data, and ultrasonic records, it would have been very difficult to rule out a casting defect. In many older mills, such records are incomplete, and the investigation remains inconclusive.
For large ductile iron castings, I always recommend that the foundry retain not only the certification documents but also separately-cast test blocks from the same melt. These test blocks should be stored in a condition that allows future testing. If a failure occurs, the retained samples can be tested independently to verify that the material delivered to the site was the same as the material that was originally qualified. This practice is particularly important for safety-critical components such as grinding table bodies, roller hubs, and other load-bearing ductile iron castings.
I also recommend that the original drawings and specifications be kept for the entire life of the equipment. The drawing should clearly state the material grade, the required nodularity, the graphite size, the mechanical properties, the ultrasonic acceptance level, and any special requirements for the casting of large section ductile iron. This information is essential when failure analysis is performed later.
Considerations for the Repair Welding Procedure
Welding repairs on ductile iron castings are always risky if the procedure is not rigorously controlled. I used a hot welding approach with a preheat temperature of 500°C to 700°C. This high preheat range is necessary for thick-section castings because the surrounding metal must remain hot enough to slow the cooling of the weld and the heat-affected zone. The weld metal from an EZNiFeCu electrode remains austenitic at room temperature, which gives it good ductility and resistance to hydrogen-induced cracking. The copper addition improves the strength of the nickel-iron matrix and reduces the difference in thermal expansion between the weld and the base metal.
The welder must be qualified specifically for cast iron welding. The process should be performed in a clean, dry environment. The welding area should be protected from draughts. The operator should monitor the preheat temperature continuously and reheat the component if necessary between passes. The number of passes should be limited, and the heat input per pass should be controlled. Excessive welding heat can cause the graphite in the heat-affected zone to dissolve and then re-precipitate in the form of carbides, which are hard and brittle. This would make the repaired area more prone to cracking under cyclic loads.
After each pass, peening of the weld bead can help relieve contraction stresses, but peening must be done while the weld is still hot and only on the weld metal, not on the base metal near the heat-affected zone. I did not consider peening to be essential for this repair because post-weld heat treatment was planned. However, if post-weld heat treatment were not possible, peening would have been a useful alternative stress-relief measure.
The post-weld heat treatment was performed at 550°C to 600°C. This temperature range is below the transformation temperature of the ferritic matrix, but high enough to relieve residual stresses. The cooling rate was limited to less than 30°C per hour. This slow cooling rate is critical because a faster cooling rate would create new thermal stresses. The success of the heat treatment was verified by hardness testing and liquid penetrant inspection.
Monitoring the Repaired Area in Service
After the mill was restarted, I advised the plant to monitor the repaired area carefully. The simplest monitoring method is periodic visual inspection of the outer surface and the upper machined face. A crack that has begun to grow will usually show a thin dark line or a small amount of fine powder caused by fretting between the crack surfaces. The plant should also monitor vibration signals. An increase in the amplitude of the vibration at the frequency corresponding to the disc body rotation can indicate a change in stiffness or mass distribution. The mill’s bearing temperature and hydraulic pressure should also be recorded because a growing crack can cause the grinding force distribution to change.
For quantitative monitoring, ultrasonic thickness and crack-length measurements can be made at the same grid points during each scheduled inspection. The measured crack length should be plotted against the number of operating days. If the crack length remains constant over several inspections, the repair is behaving well. If the crack length increases, the mill should be stopped immediately. The table below gives an example of the monitoring schedule I recommended.
| Inspection interval | Inspection method | Acceptance criterion |
|---|---|---|
| Every week | Visual inspection around the repaired area | No visible crack growth, no loose weld, no spalling |
| Every month | Liquid penetrant testing on the repaired surface | No surface cracks |
| Every three months | Ultrasonic inspection of the repaired section | No indication exceeding the original Level 2 acceptance limit |
| Every six months | Full review of mill vibration, temperature, and pressure records | Vibration within normal range; no abnormal hydraulic pressure change |
Long-Term Replacement Planning
No welded repair of a heavily loaded ductile iron casting can be considered a permanent solution. The heat-affected zone and the weld metal will never have exactly the same mechanical properties as the original casting. The repaired area will also have a different residual stress distribution than a virgin casting. Therefore, I recommended that the plant order a replacement disc body as soon as possible. The replacement should be manufactured according to the original drawing, but I recommended several improvements for the long term.
First, the transition radius between the outer lifting lugs and the upper face should be increased as much as possible. A larger radius reduces the local stress concentration and makes the component less sensitive to thermal fatigue. Second, the casting design should include smooth changes in section thickness to avoid local hot spots during solidification and during service.
Third, the replacement material specification could be evaluated against an alternative ductile iron grade with improved high-temperature strength. For example, some silicon-molybdenum ductile irons have better elevated-temperature strength than ordinary ferritic ductile iron. However, the final choice must also consider machinability, cost, and the availability of foundry qualification data. For this application, respecting the inlet temperature limit is more effective than changing the material grade.
Finally, the new disc body should be inspected with the same ultrasonic acceptance level and should be supplied with full material documentation. I also recommended that the owner perform a baseline finite element analysis for the new design under the actual operating temperature, including the effect of thermal loading. This would provide a much clearer picture of the true combined stress level than a room-temperature mechanical analysis alone.
Lessons Learned
This failure contains several important lessons for all operators and maintenance engineers working with vertical roller mills and ductile iron castings.
The first lesson is that a material can be perfectly acceptable according to its certification and still fail if it is used outside its intended thermal envelope. EN-GJS-400-18 is an excellent material for grinding tables at moderate temperatures, but its strength drops rapidly above 400°C. The customer must know this limit and must ensure that the process control system cannot exceed it, even during upset conditions.
The second lesson is that emergency cooling methods can cause more harm than the normal operation itself. A plant that wants to enter the mill quickly may be tempted to blow compressed air onto the hot disc body. This action can create a thermal shock that cracks a large casting. The correct approach is to cool the mill naturally or to use the main gas fan with controlled temperature ramping.
The third lesson is that the location of a crack can provide valuable diagnostic information. In this case, the crack started at a geometric stress concentration and propagated along a plane of maximum thermal stress. By reading the crack path, I was able to confirm the root cause and avoid wasting time on complex metallurgical investigations of an already acceptable casting.
The fourth lesson is the value of a comprehensive repair procedure. Repair welding of ductile iron castings requires careful attention to preheat, electrode selection, interpass temperature, post-weld heat treatment, and non-destructive testing. A repair performed without these controls would almost certainly fail quickly.
The fifth lesson is that the repaired component must be considered a temporary solution. It may run safely for a short period, but it will never be as good as a new casting. The owner must use the time after the repair to plan for a permanent replacement and to implement operating changes that prevent the same failure from recurring.
Summary of the Repair Outcome
The repair was completed according to the procedure described above. Crack-stop holes were drilled, reinforcement plates were installed, the welding was performed with the EZNiFeCu electrode, and the component was post-weld heat treated and inspected. The mill was then restarted and operated under the revised thermal controls. The disc body remained stable during the temporary operating period, and the scheduled inspections did not show any new crack growth. The new spare disc body was ordered, and the mill was later shut down for replacement. The repaired disc body was retired from service without any unsafe incident.
The success of this repair depended on three factors. The first was the accurate root cause analysis, which identified the true problem as thermal overload and not material quality. The second was the discipline of the welding team in following a careful preheating and post-heating procedure. The third was the cooperation of the plant operators in respecting the new temperature limits and avoiding rapid cooling after shutdown.
In the final analysis, the most important message is that large ductile iron castings must be treated as engineering components with real limits. They cannot be exposed to any temperature simply because they are large and heavy. The safe operation of a vertical roller mill depends not only on the strength of the design but also on the ability of the process to keep the hot gas temperature below the limit that the material can withstand. This case demonstrates that with a proper failure investigation, a controlled repair, and a strong operating discipline, a serious cracking incident can be resolved safely and economically.
Conclusion
The failure of the ductile iron disc body was not caused by a casting defect, by insufficient mechanical strength, or by an error in material selection. The root cause was thermal overload. The inlet gas temperature exceeded the safe working temperature of EN-GJS-400-18 ductile iron, and the use of compressed air for rapid cooling created severe thermal fatigue. The combined effect of reduced high-temperature strength, thermal stress, phase transformation stress, and geometric stress concentration caused the disc body to crack at the outer lifting-lug region.
The most important conclusion for plant operations is that the temperature of the hot gas entering the mill must be controlled so that the metal temperature of the disc body remains below 400°C. At temperatures above 400°C, the strength of ductile iron castings falls rapidly, and the safety margin that exists at room temperature can disappear. If the process requires higher temperatures, the design must be changed, or a different material with improved high-temperature capability must be selected.
For the long-term reliability of vertical roller mills, I recommend combining robust design, high-quality ductile iron castings, complete material traceability, finite element analysis under both mechanical and thermal loads, and strict process control. The repair described in this paper gave the plant a safe window of operation, but the permanent solution lies in respecting the material limits and replacing the damaged component with a properly manufactured casting. The analysis and repair procedures presented here can be used as a reference for similar failures in ductile iron castings.