The Research Of Acoustic Signal Recognition And The Application On Machine Running State Forecast

As the developing of the computer technology, voice pattern recognition technology basing on acoustic signal processing grows fast in recent years. Acoustic signal is stability and accuracy , so using acoustic signal to classify objects can expand the field of pattern recognition .Sound, which the machine'signal when it is running and sending out. it must contain the machine's plenty of structure information and statue information .Through it can predict the machine's running statue and perform inspect on line. So voice pattern recognition would be one of way to machine's inspection on line and fault-forecast. It will be main direction of fault-prediction filed's development.The electrical engineering used for driving kinds of machine nowadays with the most in general devices in industry, for guaranteeing its safe and steady circulation, usually needing the workers' periodical checking and maintenance. While the breakdown of electrical engineering takes place, the workers judge the breakdown type by hearing the voice of electrical engineering sending out. However, this kind of artificial method not only waste the manpower of large quantity, but also can't guarantee breakdown removing on time . Making use of the computer to proceed engineering's sound could judge the type of fault running-time instead of the workers, it would save much manpower. According to this kind of thinking, in this paper theories of pattern recognition would be applied to the automatic examination system in electrical engineering running-time.The main content in this paper is the following aspects:First research several ordinary fault type in machine running,through analyzing them to put a project as a whole and technical route of machine's running statue and forecast, think of traditional Boolean logic recognition and FAT is not correct ,can't meet require, in this design use method of wavelet packet analysis; Extract character from feature signal ,choose sensitive character parameters;Second research application technology of BP neural network on machine 's statue prediction, in allusion to the current neural network's fault, afford a kind of new type assorted neural network to fit for the require of prediction; To research the machine's statue and the technology ,using trinomial asynchronous electromotor of Y630-10/1180 to study in locale, it can recur the faults' development and tendency, and analyze these results.