Research On Mechanical Fault Prediction Based On Full Vector Support Vector Regression

In order to meet the needs of the rapid development of society, mechanical equipment in the first guarantee the production quality and reliability of the equipment under the premise in order to improve the efficiency is more and more high-speed, in order to reduce labor costs, more and more intelligent, accompanied by the development of the equipment becomes more and more complex. Rotating machinery is as the core equipment of machinery and equipment, operation and efficient operation and reliability and the overall industrial development is closely related. Followed by the rotating machinery and equipment structure is more complex, its running state is more and more complex, more and more attention to the equipment operation and maintenance personnel. Usually a failure occurs in light of the impact of processing quality and equipment stability, weight is often accompanied by casualties, property damage and production disruption. It has become an important subject in the industrial technology for how to make the equipment run safely and reliably. Not only to real-time monitoring of the current running state of the equipment, but also according to the current state of the future state of the correct forecast in a timely and effective analysis of the failure process, the failure to resolve in the bud. Can also be combined with the operating rules of the previous machine to arrange the maintenance time, reduce production costs. On the spectrum of the fault of prediction can be more accurately predicted that failure, can be targeted to exclude some unrelated fault, the focus lock observation and solve the fault, can the fault more accurately judge the processing, greatly reducing maintenance time and maintenance costs, improve equipment utilization rate, safety and reliability, avoid the occurrence of disastrous accidents, for the normal operation of the equipment significance major.The shortcomings of the full vector spectrum technology from the single channel data information one-sided incomplete, the multi-channel data of information fusion and the vibration information of the rotor are accurate and comprehensive extraction fusion, can truly reflect the running state of the equipment. In this paper, the trend prediction of rotary machinery on line monitoring and fault diagnosis system is studied, and the full vector spectrum technology is applied to the trend prediction and fault diagnosis of rotating machinery. The main research work is as follows:1. This paper expounds the basic theory of the full vector spectrum technology, its advantages in equipment vibration signal processing and the accuracy of the fault diagnosis. It shows that the full vector spectrum technology has advantages in processing the vibration signal in equipment operation.2. This paper describes the basic theory of support vector regression, numerical algorithm, and the data do trend prediction, prove the role of support vector machine in fault prediction.3. The basic theory of the new method based on the support vector regression(FVSVR) and the accuracy and validity of the new method in the prediction of data are expounded. Through the analysis of the actual data, using the method, the single value prediction frequency multiplier is used to predict the spectral structure. It is proved that the method is effective for the prediction of the vibration signal spectrum of the running state of the equipment.4. Using the full vector support vector regression to predict the spectral structure of the vibration signal of the equipment operation, and the accuracy and practicability of the method is verified by an example. The influence of the different frequency harmonic prediction parameters on the prediction accuracy is discussed, and the parameter selection of the spectral structure prediction by using the vector prediction is established.