Decision support algorithm to prevent hardware failures

Abstract

Diagnostic systems implemented at many industrial facilities are designed to detect defects in a timely manner, thereby prevent sudden failures with subsequent equipment downtime, accompanied by economic losses. Most diagnostic systems are based on making a decision about the serviceability or malfunction of an object based on comparing diagnostic parameters with threshold values. The problem with this approach is that going beyond the set limits is fixed after the equipment has lost maintainability or requires complex expensive repairs. Therefore, the goal is early and reliable detection of defects in electromechanical equipment. The goal is achieved by developing an algorithm for detecting the beginning of a deviation in a time series consisting of sequentially recorded diagnostic parameters. The algorithm is based on the consistent application of singular spectrum analysis and allows not only timely detecting of deviations, but also assessing the probability of a decision error. Data processing algorithm is investigated on data sets generated in accordance with the specified laws of change. Moreover, the possibility of timely and reliable detection of defects has been demonstrated on a real set of diagnostic data. The implementation of the algorithm in the management of equipment condition will contribute to improving the safety and efficiency of production facilities.

Keywords

• diagnostic system
• vibration control
• detection of changing points
• state management
• decision errors
• singular spectral analysis