Signals Analysis Method Based On Artificial Olfaction For Equipment Fault Detection

In real life, industry and agriculture, many product quality inspecting depend on man's olfactory, including wine, tobacco, cosmetic, beverage and so on.. However configuration and function of man's olfactory are limited for some smell, it doesn't recognize these smell well. Artificial olfactory and electronic nose are popular technology in recent years. It can recognize smell well,which combines sensor array, signal processing and pattern recognition to simulate man's olfactory. Artificial olfactory has great effect and foreground in industry and agriculture, especially in equipment fault detection.Signal processing and pattern recognition which recognizes smell are important parts of artificial olfactory. At present pattern recognition technology is composed of some traditional methods or combine of some methods. Support Vector Machine (SVM) method fewer applies to artificial olfactory, especially equipment fault detection. SVM is a new classification and regression algorithm, it has great advantage, including solving small samples learning problem, having good general ability, solving excessively learning problem and partial minimum problem in some degree. So applying SVM to equipment fault detection based on artificial olfactory has many help. It can preferably draw useful information of data to enhance accurate rate of equipment fault detection.For this reason, focused on equipment fault detection, the analysis method of equipment fault signal is studied. The main contents of the thesis include:(1) The background and meaning of artificial olfactory in equipment fault detection are studied, the present condition of pattern recognition method on artificial olfactory and SVM are introduced in detail in the thesis.(2) The principle and algorithm of Principal Component Analysis(PCA) and Partial Least Square(PLS) method are introduced, the two algorithms are verified by liquorices samples.(3) The principle and algorithm of SVM is introduced. The SVM is verified by classify the principal component of liquorices, and the samples can be preferably classified.(4) The diesel oil and engine oil collected by oil electronic nose system are verified by PCA and SVM. The result shows it has a good performance.As a result, in equipment fault detection based on artificial olfactory, applying PCA and SVM to the signal analysis has good effect and foreground.