Improving the prediction of condition using simple sensors

Abstract

One approach to condition monitoring of industrial equipment is to install special purpose sensors to monitor the quantity of interest. The use of such sensors may complicate the installation of the equipment, add significantly to the cost or even depress reliability. Most modern industrial equipment is equipped with sensors for control purposes. Usually such sensors are specified for low cost and robustness and are read as part of the control process. A monitoring system which can make use of such sensors to extract detailed information adds very little to total system cost. Estimation methods offer the possibility of extracting system information from sensor signals. The monitored equipment is considered to be made up of a series of processes each with an input and output. Each input/output pair may be modelled. It is these models which permit comparison between “healthy” and current conditions. Reported results from the method are excellent and will be reviewed. In recent project work Cambridge Consultants Ltd has demonstrated that the techniques are applicable to diesel engines and provide a good initial diagnosis of a variety of faults.