Research On The Gas Flow Sensor Based On Hot-film Probe

With the improvement of society and progress of science and technology, the application of flow measurement has already penetrated into all the fields of people’s lives deeply. For instance, oil sanctions, natural gas trade and such on; the use of coal gas, the air intake control of an engine in a car and such on; when it comes down to chemical industry, in chemical reactions in which fluids participate or are produced, flow measurement is indispensable. All in all, flow measurement involves in all aspects of people’s lives and is closely related to the national economy, the building up of national defense, scientific research. Flow measurement is a detection technology of great practicality. However, there are a great deal of problems because of the complexity of the fluids. Doing a good job in the flow measurement is of great importance in guaranteeing the quality of products, improving production efficiency, developing science and technology. In particular, in today’s era while there is energy crisis and automation degree of industrial production becomes more and more advanced, flow meters are more and more important in national economy.In the research, a Wheatstone bridge is built with FS5which is constituted of two PT resistors. Based on the bridge a thermal flow sensor is produced with other digital signal processing circuits. At last the researchs on the circuit system and characters of the output curve are done.In order to solve the disadvantage that fitting accuracy in high flow fluid and that in low flow fluid can’t be given consideration to at the same time, the importance of the piecewise fitting method is analyzed in three aspects including the results of traditional polynomial fitting, the physics principle analysis of the thermal transmission of the thermal probe and the analysis of others’ data. A preliminary systematic research on the piecewise fitting method of which the advantages are pointed out has been done, some related demonstrations are shown.At last, the support-vector-machine(SVM) algorithm is introduced into the fitting of the output curve of the thermal flow sensor. According to the data the parameter of kernel function is optimized, realizing the high-accuracy fitting of the variables, expecting that the whole fitting process can be accomplished automatically and intelligently. With the comparsion of the result of the SVM algorithm and the result of the traditional polynomial fitting, the advantages and disadvantages of the intelligent algorithm are demonstrated.