Study On Theory And Application Of Intelligent Condition Monitoring For Compressors

Compressors are applied in every walk of life more and more widely with thedevelopment of technology of automation. If compressors appear abnormal condition orsudden fault in the process of running, it will bring on product lines and equipmentsstopping, personnels of enterprises hurt or even exploding and thus produce greateconomic loss. It is necessary to do a research with condition monitoring methods andtheories of compressors and build a perfect condition monitoring system of compressors.In this dissertation, intelligent condition monitoring theories and implementationtechniques are systematically researched combining with the actual need of a certaincompressor company.This dissertation firstly makes theoretical research of compressors' working principlesand their basic structures, analyzes compressors' typical fault modes, analyzescompressors' condition signals and then contact compressors' fault modes with conditionsignals in detail. These works establish theoretical foundations for the best research oncompressors' condition monitoring theories. For providing a realization way, acompressors' condition monitoring trial platform is designed.Intelligent condition forecasting models of compressors are built. After analysis oftheir characteristics, this dissertation brings forward methods of adopting two kinds of BPnetwork forecasting model and ARIMA assembled forecasting model to monitorthermodynamics parameters and electric parameters of compressors. These models aretested and applied in a compressor running field. The wavelet theories are researched. Thisdissertation brings forward applying binary wavelet transform to extract condition featuresof the pressure pulsing signal. The feature parameters used to show compressors' workingcondition—wavelet decomposed high frequency coefficients are obtained. The methodprovides a new way for monitoring compressors condition.Principles of eliminating noises from vibration signals and arithmetics of vibrationparameters conversion are researched. After that, the method of eliminating noises withwavelet packet and the method of sub period integral are given to be applied inpretreatment of vibration signals. These methods improve the vibration signals' precisionin compressors' condition monitoring. This dissertation discusses methods ofcompressors' frequency response characteristic test with an application example and thendesigns a frequency response characteristic test system based on Virtual Instrument. Two new condition monitoring methods with compressors' vibration signals time-frequencyanalysis—continuous scale-wavelet energy monitoring and binary wavelet sub frequencybands energy monitoring are brought forward with practical application based ontheoretical research.Noise signals' collection methods are researched. A noise signals analyzing systembased on Virtual Instruments is also designed. This dissertation makes theoretical analysiswith noise sources and indicates adopting methods of multi-working condition test and 1/3octonal frequency spectrum to find compressors' noise distribution characteristic. Methodof recognising noise sources with continuous wavelet is also discussed. Then, theories ofBlind Source Separation are researched and applied in compressors' noise signalsmonitoring.This dissertation also builds a multi-sensor information fusion system frame forcompressors' online monitoring and fault diagnosis. Dempster-Shafer theory of evidence indeciding level fusion is researched and DSET fusion diagnosis model is adopted to realizemulti-criterion fusion technology. This dissertation especially analyzes a computingexample for the compressor information fusion.Finally, according to compressors' online monitoring development and demands, thisdissertation makes application with research results of intelligent condition monitoringtheories by adopting design thought of Virtual Instrument. So, a compressors' conditionmonitoring system are designed including the whole structure, software and hardware.Many field experiments proceedes and the satisfied results obtained shows the newmethods researched validity and practicability.