9x Common Faults Detected in Steel Manufacturing

Simon Jagers
8-minute read
Updated: 02/01/25
Published: 02/01/25

In this case study, we explain how SAM4 detects common faults in steel mills using actual results from anonymized SAM4 data, to help engineers evaluate SAM4’s value for their own operations.

Mechanical indicators

Failing components leave distinct signatures in the current and voltage signals at frequencies related to their geometry. For example, roller bearing issues can be identified by changes at the bearing’s ball pass, ball spin, and cage frequencies, which depend in part on the number of rollers and their size.

Figure 1. Spikes around the supply frequency at this bearing’s characteristic frequencies indicate developing bearing damage (blue = healthy, black = bearing damage, green = after replacement).

Sam4 dashboard analyses — Figure 2. Sample SAM4 dashboard analyses for two different mechanical faults.

Example 1: Defective coupling

Changes in the frequency spectrum led SAM4 to diagnose a defective coupling in one of the runout table rollers. We first alerted the customer at the yellow dot in figure 4. The slow rise continued, and three weeks later (orange dot) we advised the customer to inspect the coupling. They found no visual signs of deterioration, but the unhealthy changes persisted, and we put the roller on red alert approximately three months after SAM4 first flagged an increase. The customer decided not to intervene, and the defective coupling broke three weeks later. They replaced it, and intensities returned to normal.

Current frequency spectrum for the roller — Figure 3. The current frequency spectrum for the roller when healthy (blue) vs. with degrading coupling (black).

Rising intensity over three month period — Figure 4. Rising intensity over a three-month period at the second harmonic of the roller’s rotational frequency. Note the drop in amplitude after the customer replaced the defective coupling.

Example 2: Defective coupling

An increase in intensity at the rotational frequency of a runout table roller led SAM4 to diagnose a coupling fault. Figure 5 shows changes in intensity at the roller’s rotational frequency over time. The customer was first alerted to a significant anomaly at the orange dot in figure 5 and advised to inspect the coupling. The intensity rapidly rose, resulting in a red alert several days later. The customer left the asset running until the coupling broke, approximately three days after the red alert. Intensities returned to normal after the broken coupling was replaced.

Rising intensity at rollers rotational frequency — Figure 5. Rising intensity at the roller’s rotational frequency over time. Note the drop in amplitude after the customer replaced the broken coupling.

Changes in frequency spectrum over time — Figure 6. Changes in the frequency spectrum over time: healthy (blue) vs. degrading (black). The appearance of regular but transient increases in noise (green) were part of what spurred the move to red alert for this machine.

Example 3: Defective coupling

SAM4 detected an increase in intensity at this roller’s rotational frequency plus the second harmonic, as well as a rise in the noise floor. These changes were consistent with the typical pattern related to looseness in the cardan shaft coupling for this customer’s other runout table rollers. We advised the customer to check the coupling for developing looseness during the next maintenance round. They notified us that they hadn’t seen any signs of deterioration, but had replaced the defective element just in case. Subsequent data showed the values had returned to normal.

Figure 7. Rising intensity at the second harmonic of the roller’s rotational frequency over time. Note the drop in amplitude after the customer replaced the defective element.

Example 4: Defective cardan shaft

Sudden changes in the current root mean square (RMS) value prompted SAM4 to flag a potential cardan shaft problem in this runout table roller. Spectral analysis revealed a rise in intensity at multiple harmonics of the roller’s rotational frequency, as well as in the noise floor. Previous experience with this customer’s rollers indicated this was a fault that can develop quickly, so we put the asset on orange alert and recommended the customer inspect the cardan shaft as soon as possible. Three days later we went to red alert. The customer noted they were waiting for the right opportunity to replace the cardan shaft. Three days later the cardan shaft broke. The customer replaced it and values returned to normal.

Figure 8. Changes in the frequency spectrum over time: healthy (blue) vs. degrading (black).

Rise in magnitude at first interharmonic — Figure 9. The rise in magnitude at the first interharmonic (part of the noise floor).

Example 5: Bearing fault

SAM4 measured increasing intensity at the cage frequency (fundamental train frequency) for one of the bearings in the motor for this runout table roller. The load on the motor also gradually increased, without corresponding process changes. We alerted the customer to possible bearing damage (orange dot in figure 10). Four weeks later they replaced the bearing during a planned maintenance stop. The motor’s scores then returned to normal.

Load on motor increasing over time — Figure 10. The load on the motor increasing over time. The customer received an alert at the orange dot, advising inspection. After maintenance the load on the motor returned to normal.

Example 6: Bearing fault

Small spikes at the cage frequency for the motor’s NDE bearing led SAM4 to diagnose a developing bearing issue. The customer was alerted (the orange dot in figure 12) and advised to inspect the bearing at the next planned maintenance stop. At the planned stop roughly two months later, early bearing damage was confirmed and corrective action was taken.

Frequency spectrum for the motor — Figure 11. The frequency spectrum for the motor when healthy (blue) and with a degrading bearing (black). The peaks correspond to the bearing cage frequency and its harmonics.

Increasing intensity at nde bearings case frequency — Figure 12. Increasing intensity at the NDE bearing’s cage frequency (blue dots). Values returned to normal after the bearing was replaced (black dots).

Example 7: Bearing fault

SAM4 observed an increase in intensity at frequencies corresponding to the roller bearing (cage frequency plus harmonics). We alerted the customer and advised them to inspect the bearing for damage. When an opportune moment presented itself three months later, they confirmed bearing damage through manual vibration measurements. They replaced the bearing and values returned to normal.

Changing frequency spectrum for the roller over 3 months — Figure 13. The changing frequency spectrum for the roller over a three-month period: when healthy (blue and black) and with beginning bearing degradation (green). The peaks correspond to the bearing cage frequency and its harmonics.

Heat map showing changes related to bearing degradation — Figure 14. A heat map showing changes related to bearing degradation. The horizontal axis is frequency; the vertical axis is time, from oldest date at the top to most recent date at the bottom. The color indicates intensity, from high in red to low in blue. New peaks at bearing-related frequencies are clearly visible in the lower half of the graph.

Electrical indicators

Developing electrical issues instantly affect the motor’s magnetic field and leave distinct signatures in the current and voltage signals. For example, a broken rotor bar creates an electrical asymmetry in the rotor which produces a counter-rotating magnetic field, inducing stator currents at multiples of twice the slip frequency around the supply frequency.

Broken rotor bar indicator — Figure 15. Another broken rotor bar indicator: the resulting thermal rotor deformations induce spikes at the motor’s rotational frequency (arrows) modulated around harmonics of the supply frequency (vertical lines).

Example 8: Stator winding short circuit

SAM4 observed an increase at the motor’s rotational frequency, which can be indicative of several failure modes. We notified the customer and asked them to inspect the roller. They found no signs of damage. Over time the intensity also began to rise at odd harmonics of the supply frequency, and SAM4 began to observe a current unbalance that was larger than for the customer’s other motors. These indicators led SAM4 to diagnose degrading stator winding insulation and move the motor to red alert. The customer replaced the motor during the next planned stop and values returned to normal.

Motors rotational frequency shows clear rise in magnitude — Figure 17. The motor’s rotational frequency shows a clear and intensifying rise in magnitude (blue dots). Values returned to normal (black dots) after the motor was replaced. Three months from SAM4’s first alert to replacement.

Motor unbalance — Figure 18. The unbalance during the period of interest (right side of graph) was notably higher than usual for this motor, and was not reflected in the other motors in the same group.

Example 9: Rotor eccentricity

SAM4 observed an increase in intensity at a frequency corresponding to the number of rotor bars multiplied by the rotational frequency, which can be indicative of rotor eccentricity. We notified the customer and continued to monitor the asset. As values continued to rise, we moved to orange and then red alert, keeping the customer informed weekly and then daily. Five months later, the customer replaced the motor and intensities returned to normal levels.