Design Of Capacitance Tomography System And Research On Multi-stream Pattern Identification Algorithm

Electrical Capacitance Tomography(ECT)is a process tomography technology based on capacitive sensitive field.This technology uses sensors to measure the required capacitance data to detect the flow pattern of the filling pipeline,so as to meet the controllable flow pattern of the pipeline.It is of great significance to accelerate the development of industrial intelligence.As a process tomography technology,ECT has many advantages.Compaxed with traditional phase flow detection technology,ECT has the advantages of non-invasiveness,visualization,fast detection speed,high reliability and strong safety.important research directions for detection.There are still problems to be solved in this technology.In this paper,the hardware design and flow pattern identification of the ECT system are deeply studied.The main work and results are as follows:1.Design the capacitance data acquisition module with STM32F103RCT6 chip as the main control unit.The module includes a capacitance sensor,a signal generating circuit,a signal amplifying circuit,an AC micro-capacitance measuring circuit and a bias circuit.According to the principle and design principle of capacitive sensor,a 12-electrode sensor array is made;SMA shielded wire is used as the wiring,which effectively shields the interference signal and reduces the stray capacitance;design the sensor control circuit,so that the sensor shares one excitation source and capacitance measurement circuit;A capacitance measurement circuit with strong anti-interference,low drift and high sensitivity is designed.Experiments show that the circuit has good performance and the measurement results meet the theoretical requirements.2.The optimization of DBN-ELM(Deep Belief Networks-Extreme Learning Machine)network based on Spanrow Search Algorithm(SSA)is studied.The ECT flow pattern identification method based on image reconstruction has low recognition rate and complicated operation.This paper proposes a flow pattern identification algorithm based on SSA optimized DBN-ELM.The capacitance data collected by the lower computer is made into a data set,the capacitance data features are extracted through the DBN network,and ELM is added on the top layer of the DBN network to complete the identification of the abstract capacitance flow pattern data;the number of neurons in the DBN hidden layer affects the learning ability of the entire model Therefore,the SSA optimization algorithm is introduced to calculate the optimal number of neurons of the Restricted Boltzmann Machine(RBM)in each layer.Compared with other identification algorithms,the accuracy rate is improved and the required time is shortened.