Mechanical Key Component Life Prediction And Maintenance Strategy Research Based On The Improved GA-SVR

With the arrival of “the era of industrial 4.0” and “the era of intelligent manufacturing”, mechanical transmission equipment system plays a more and more important role in areas such as aerospace, marine and high-end manufacturing. So, the reliability of equipment system and certain key components are regarded as one of the important concerns of equipment health management. Remaining useful life prediction and maintenance mode decision are not only two extremely important aspects of equipment health management, but the important means to improve the reliability of the equipment. By using scientific prediction method to promote life prediction accuracy and choosing the appropriate maintenance strategy is a very effective way to promote the reliability of the equipment system and its key components on the basis of the forecast results.This paper considered two aspects which respectively are the research and improvement of prediction algorithm and maintenance strategy decision. By using a certain method of data mining and intelligent machine learning algorithms to promote the accuracy and efficiency of life prediction. And then, the maintenance cost model considering reliability is established to select the optimal maintenance mode and maintenance time. And the purpose of this model is to promote the reliability of key components and the economy of the health management maintenance which has important academic value and practical significance. And the main research content of this paper is as follows.(1) This paper analyses the research status, main problems and the current research hotspots and difficulties both at home and abroad of the remaining life prediction technology. And summarizing the theoretical system, the main influence factors and the future research trend of maintenance decision-making, then concluded the research problem, objectives and significance of this paper.(2) To process and utilize large-scale health data of real-time online, this paper proposed the remaining useful life prediction model of GA-SVR to mechanical key component’s large-scale health status data problem, which based on fusion increment LHD sampling algorithm and introducing the quadratic exponential smoothing to preprocess before the input of the index set of performance decline to improve the model prediction accuracy and generalization performance. Compared with the standard SVR model, the improved increment GA-SVR prediction model presented a distinct advantage in the aspect of prediction accuracy and prediction efficiency by the verification analysis, which the prediction error dropped by 44.89% and the prediction time reduced by 39.87%.(3) This paper established an optimization model of the maintenance strategy decision, which oriented to the zero unscheduled downtime and aimed to minimize the per-unit maintenance cost of mechanical key components within their life cycle. And the model considered the change of the failure probability with the change of the service time and the effect of the maintenance cumulative damage in advance on the maintenance cost and introduced the concept of maintenance time threshold. This model will eventually give the optimal way of maintenance and the optimal point of maintenance time. Finally, the experimental example proves the validity and feasibility of the models.