Research On Equipment Fault Prediction Based On Information Fusion

Mechanical equipment has played a huge role in promoting industrial progress. In order to ensure the continuous and efficient operation of key equipment in industrial production, and improve productivity, accurate diagnosis and prediction for equipment fault has gradually become the most important of the industrial development.Most of the fault prediction techniques in the past are based on single-sensor signal, however, due to the eddy whirling of rotor, the signal collected by a single sensor is often one-sided, which leads to inaccurate prediction result. While the full vector spectrum technology which can integrate multi-channel signals and extract characteristics accurately, overcomes the technical difficulty of unreliable prediction data and improves the accuracy of fault prediction effectively.The support vector regression has obvious superiority in solving the small sample prediction problem, and the least squares support vector regression has better generalization ability and prediction performance which expanded on the basis of it. Using least squares support vector regression in fault prediction can make full use of small sample data collected recently from equipment condition monitoring system to predict the trend, and this will greatly increase the credibility of prediction result, with good engineering practical effect.In this paper, the full vector spectrum technology is combined with the least squares support vector regression method, to construct an equipment fault prediction model based on information fusion. Different from previous time domain index prediction, this model is to predict the spectrum structure which reflects the operation state of the equipment, and its effectiveness is verified by an experiment. The main research work of this paper is as follows:1. Analyzing different information fusion methods and the advantage of full vector spectrum technology is illustrated through comparison. The principle, algorithm and characteristics of full vector spectrum technology are introduced, and it can be proved that full vector spectrum can extract complete and accurate feature information from the same-source signal through an example of fault diagnosis on Wind Turbine Drivetrain Diagnostics Simulator(WTDS). By using the full vector spectrum technology, the characteristics of the real equipment state can be obtained, which provides reliable data support for fault prediction.2. Study the basic theory of support vector regression, and analyze the idea of regression prediction, then construct the support vector regression prediction model based on full vector spectrum. This model solves the problems that the previous single-channel data is incomplete and the prediction result is inaccurate, and its feasibility is verified through the sinc function simulation and steam turbine unit 1X prediction example.3. Construct the full vector least squares support vector regression prediction model, and give its specific algorithm and prediction flow. By applying the actual vibration data of the steam turbine unit to this model, it is proved that the model can predict the spectrum structure and get a better prediction effect, with a very good engineering practical significance.