The Research Of Transducer’s Consiscency Based On Principal Component Analysis’s Genetic Neural Network

Gas ultrasonic flowmeter first appeared in the 1970s, which has many outstanding advantages, such as stable and reliable performance, wide range ratio, no pressure loss of detection and so on. The most used measurement method is based on velocity difference of time difference. What’s more, the ultrasonic transducer used in this method is the most basic and important sound electric energy conversion sensor, whose performance directly affects the instrument’s overall performance. People do little study of the paired technology of transducer, and the most technology still stays in practical experience. To deal with the condition, this paper put forward a paired technique which is based on component analysis’s genetic neural network. This paper is divided into six chapters:The first chapter of the thesis summarizes the gas ultrasonic flowmeter and the ultrasonic transducer, and introduces the gas ultrasonic transducer paired significance and method, besides, it emphatically introduces the problem of paired, so as to put forward transducer paired technology research in this paper.Paper in the second chapter builds equivalent circuit model of the transducer, and establishes the correlation equivalent circuit model, studys the sensitivity of the transducer and the electrical parameters through the equivalent circuit model afterwards, then carrying out experiment to verify the above theory, and finally gives the relationship between the performance parameters of the transducer, which provides guidance for the predictionThe third part studies the basic principle of neural network algorithm, because the network input layer has too many parameters, so using principal component analysis to remove the correlation among those parameters. And to solve the problem existing in the neural network, this paper makes the use of genetic algorithm to optimize weights and valves of the traditional BP network, the precision of model is improved by solving these two problems, and finally builds transducer paired model based of analysis’s genetic neural network.The fourth part uses the simulation data to verify the accuracy of the model, and compares the prediction precision of different models to get the best prediction model, which is the genetic neural network model based on principal component analysis, then uses this model to forecast the actual data in different conditions. At the end, compares the simulational prediction results with the experimental results.The fifth part designs the flow chart of the optimized paired program and performance prediction program, and using the MATLAB GUI to design the paired systerm of the transducer.Thesis finally summarizes the research work of this project, and combines the actual situation to put forward some prospects.