Multisensor Condition Monitoring and Anomaly Detection in Conveyor Motors: A Retrospective Analysis of Temperature, Vibration-Related Signals and Operating Speed

Abstract

Conveyor-motor condition monitoring combines measurements that may reflect thermal state, motion and operating behaviour, but retrospective sensor records can be difficult to compare across sessions. This study analysed archived records from conveyor Motors 3 (M3) and 4 (M4), including normal-operation, recorded-stall and pallet-static/potential-fault sessions, together with separate speed and temperature/vibration-related files. Valid observations were summarised into one-minute windows for descriptive analysis, timestamp alignment and exploratory anomaly detection. The median of one-minute temperature medians was 23.89 °C in the M3 normal record and 31.34 °C in the M3 recorded-stall record. Corresponding summaries were 39.38 °C for M4 recorded stall and 39.71 °C for the M4 pallet-static/potential-fault record. The separate M4 speed file contained 3,781 valid readings, of which 1,415 (37.4%) were zero. The supplied file review indicates that 1,408 zero readings occurred in a sustained end-of-record block, while the remaining seven zero readings were associated with brief dips. These speed patterns do not independently confirm mechanical stalls because synchronised motor-current measurements or independent event logs were unavailable. The separate speed stream was reported at approximately 1 Hz, while the temperature/vibration record was reported at approximately 100 Hz. An Isolation Forest trained on the M3 normal session flagged all windows from the other records, indicating differences between recording sessions rather than verified fault-detection performance