Failure mechanisms discrimination and life prediction of safety relay

The failure mechanisms of a safety relay were analysed and divided into three failure modes. The contact resistance, closing time, super-path time, bounce time, releasing time and arc time were used as failure characteristic parameters, and the wavelet transform and moving average methods were utilized to achieve denoising. Aimed to the possibility of correlation among characteristic parameters, the principal component analysis method was used to achieve the dimension reduction for the multi-dimensional characteristic parameters to correlate the characteristic parameters. The average gradient value of principal components was defined. The failure modes of 18 sets of contacts were discriminated by using the Euclidean distance discrimination criterion. The characteristic parameters of contacts were analysed using the Fisher discrimination method. The characteristic parameter significantly reflected the features of a safetyrelay's life was selected as the prediction parameter. A grey prediction model was established to predict the safety relay's life, and the prediction effect of the model was verified. Research result indicates that after denoising, the characteristic parameters have good smoothness, high signalto-noise ratios, and low root mean square errors, indicating that a good denoising effect is achieved. The failure modes of 18 sets of contacts are discriminated with an accuracy of 83.3% using the Euclidean distance discrimination criterion. The super-path time has the largest Fisher discrimination function value of 8.2, indicating that the super-path time is an important parameter to represent the life of a safety relay. The actual lifetime of the safety relay is 122 000 times, the prediction life is 116 000 times based on the grey model, and the relative error of the prediction life is 4.9%, indicating that the proposed life prediction model has high accuracy and good feasibility.