Rotating electrical machines permeate all areas of modern life at both the domestic and industrial level. The average modern home in the developed world contains 20-30 electric motors in the range 0-1 kW for clocks, toys, domestic appliances, air conditioning or heating systems. Modern cars use electric motors for windows, windscreen wipers, starting and now even for propulsion in hybrid vehicles. A modern S-series Mercedes-Benz car is reported to incorporate more than 120 separate electrical machines.

It is necessary to identify the condition monitoring that needs to be applied to detect those faults covered in the previous chapter. Condition monitoring has been seen primarily as the province of those who analyse fault signals and interpret results. However, interpretation must be informed by causality and the authors' view is that one of the important changes to occur in condition monitoring over the last 30 years is ensuring that it is firmly directed towards the root causes of machine failure. In so doing, successful condition monitoring directly addresses machine unreliability or, more importantly for the operator, lack of availability. This chapter sets out the issues concerned with this causality.

This chapter describes the signal processing methods now available for computer monitoring electrical machines. The following five chapters describe the major temperature, chemical, mechanical and electrical techniques available for monitoring.

This chapter shows that chemical and wear analysis have been demonstrated to be effective global monitoring techniques for electrical machines which can produce narrow bandwidth (<1 Hz) signals. Chemical degradation of insulating materials and lubricants are detectable and can give bulk indications of the condition of an electrical machine. This is particularly important when it is considered how central insulation and lubrication integrity are to the long-term life of a machine. However, the detectability criteria for these techniques are difficult, the chemical analysis processes involved are complex and expensive, and the quantity of data generated by chemical analysis currently confine their application to only the largest machines.

This chapter shows that electrical techniques are powerful tools for the condition monitoring of electrical machines, particularly axial leakage flux, current and power, offering the potential to provide a general condition-monitoring signal for the machine. The availability of high-quality, digitally sampled mechanical vibration and electrical terminal data from electrical machines opens the possibility for more comprehensive monitoring of the machine and prune mover or driven machine combinations. However, these signals generally require broad bandwidth (>50 kHz) and a high data rate for adequate analysis. Therefore the principal difficulty of applying these techniques is the complexity of the necessary spectral analysis and interpretation of their content. This situation is made more difficult if variable speed drives are involved because tune domain signals may no longer be stationary and will also be polluted by harmonics from the power electronic drive. Comprehensive monitoring of an electrical machine can be achieved by measuring shaft flux, current, power and electrical discharge activity. These are broad bandwidth (generally >50 kHz) signals requiring complex analysis. Shaft flux, current and power signals are capable of detecting faults in both the electrical and mechanical parts of a drive train. Shaft voltage or current is an ineffective condition monitoring technique for electrical machines. Shaft flux monitoring is non-mvasive and uses a single sensor but it is complex to analyse and untested in the field. Current monitoring is also non-mvasive, but uses existing sensors and has established itself as motor current spectral analysis, a reliable and widely accepted technique for machine monitoring. Power monitoring is also non-mvasive, uses existing sensors but requires less bandwidth (<10 kHz) and less complex spectral interpretation to detect faults but is not yet widely accepted, so it deserves investigation for future development.

The techniques of artificial intelligence have been seen as potentially valuable for the condition monitoring of electrical machines because the underlying physics of machine operation and dynamics, as shown in Chapters 8 and 9, is so rich in fundamental rules, and these could be exploited by an expert system. This chapter shows that these techniques can be used to improve the signal-to noise ratio in a fault situation, particularly in the complex area of monitoring terminal conditions and the difficult subset of that, the detection and measurement of partial discharge activity. However, the development of artificial intelligence for electrical machine condition monitoring is still in its infancy and despite the considerable work that has been done in this area, much more is required to bring such techniques into the mainstream of condition monitoring. An abiding lesson from this chapter is to understand the importance in condition monitoring of relating various monitoring signals with one another, what was described in the first edition of this text as multi-parameter monitoring. It is clear, however, that the future of machine condition monitoring will be heavily affected by multi-parameter monitoring and by the application of artificial intelligence to that process.

This appendix presents the machine structure, the failure mode and root causes in rotating electrical machines which are shown in figures 2.10 and 3.7.