Feasibility assessment of condition-based monitoring of grain silo equipment

Abstract

Maintenance is a crucial support function within the grain-handling industry. Advances in maintenance have led to the introduction of improved maintenance techniques, such as preventive maintenance, which includes condition-based and predictive maintenance strategies. Therefore, the primary purpose of the study was to determine the feasibility of implementing conditionbased monitoring and its supporting technologies on equipment typically found in a grain silo plant. The successful implementation on a small scale showed the benefits of this technique by providing information that improved the overall maintenance strategy, proving that maintenance decisions could be driven by data rather than waiting for the equipment to fail before acting. Current maintenance strategies used in industry were evaluated to identify the most appropriate condition monitoring technology and strategy for implementation in the specific environment of a grain silo. Specific focus was given to the induction motors driving the grain-handling machines, establishing how the failure modes of these equipment can be detected using the selected condition monitoring technique. An in-depth criticality analysis and review of historical maintenance data from a case study silo plant established the methodology with which to evaluate the feasibility of condition-based monitoring. Suitable methodologies found in the literature provided the framework for the development of a practical health index (HI) model. An extensive experimental method addressed a common gap in both the literature and the theory, that is, the lack of healthy vibration profile data for Class I induction motors, rated from 0.55 kW to 15 kW. A total of 2660 vibration measurements were collected from 98 different loading and operation configurations of new Class I induction motors, creating a comprehensive database of healthy vibration data. The data and subsequent regression analysis showed a correlation between the vibration amplitude (mm/s) of the rotational frequency component of healthy Class I induction motors and their power ratings (kW). Analysis showed that a linear regression equation, together with 95% confidence bounds, would be able to accurately estimate the healthy vibration amplitude of the rotational frequency for these motors. A generation algorithm for synthetic healthy signals used the regression fit results to generate a set of 2000 artificial vibration signals for each motor size. A robust Health Index (HI) model, based on the JPCCED-HI method, was developed and trained on the healthy data set. A diagnostic algorithm was added to the HI analysis framework to identify the most common failure modes of induction motors. The implementation of the HI model was proved successful by monitoring the condition of ten critical system motors in the silo plant during operation and diagnosing probable faults in three of these motors, which were subsequently confirmed by inspection and thus validated the approach. The implementation of the HI model revealed that the industry partner lacked definitive data for maintenance decisions and was unaware of the condition of the equipment. Therefore, it could be concluded that the implementation of condition-based monitoring for grain silo equipment is feasible and beneficial by providing actionable information that improves maintenance decision making.