Research On Data Processing Method Based On Extension Neural Network And Compressed Sensing And Its Application

In the modern industrial process,data generated by testing and condition monitoring of complex system is growing exponentially.It is very important to effectively deal with these data.However,because useless information will produce information distance(Information-state Transition Distance,DIST)in data processing,it is very easy to lead to “rich information and poor knowledge”.Considering this problem,people want to have an in-depth analysis of these data and find the hidden rules in it,and then extract effective information to play the “raw materials” role in the industrial field.The research directions of data processing and analysis have a lot,such as classification,clustering,estimation prediction and correlation analysis.In this paper,for the demand of petrochemical enterprises in the production process,we use extension neural network and compressed sensing theory as a research tool to study the data processing and analysis methods.The main work is as follows:1.To accurately and quickly identify the potential risk status patterns of flue gas turbine,a novel potential risk status pattern recognition method based on differential evolution-extension neural network(DE-ENN)is proposed.In the paper,through extension theoretical analysis for operation of flue gas turbine,the matter-element model of flue gas turbine is firstly established to determine the feature vectors and potential risk levels.Secondly,the differential evolution theory is led into extension neural network to propose a solution to automatically tune the learning rate and weighted coefficients.Besides,some data sets from UCI are tested to verify the effectiveness of the proposed method.Finally,the DE-ENN is applied tothe potential risk status pattern recognition of flue gas turbine.The experimental results show that the proposed DE-ENN has the advantages of less learning time,higher accuracy and less memory consumption.Meanwhile,the results also show that the method has a better performance in generalization ability and fault-tolerant ability.2.To solve the classification problem of complicated sample data,a novel extension neural network(ENN)classification algorithm based on margin discriminant projection and improved semi-supervised affinity propagation(MDP-ISAP)is proposed.At first,the margin discriminant projection(MDP)method is used to reduce the dimensions of initial data and extract the key features.Secondly,clustering analysis is conducted in the dimension-reduced feature space.On one hand,the affinity propagation(AP)clustering method is used to select sufficient training samples;on the other hand,the ISAP clustering method is used to obtain class centers which is used as the initial cluster centers,and then a novel ENN classifier is constructed.The simulation demonstrates the effectiveness of the proposed algorithm.Finally,the proposed method is employed to the High Density Polyethylene(HDPE)in the complicated chemical process.The results also show that,compared with the ENN,the proposed algorithm has a better performance.3.To improve the ENN's ability to fault diagnosis in the complicated chemical process,which has some features such as high dimension,nonlinear and non-Gauss distribution,a novel MultiBoost-based integrated ENN(extension neural network)fault diagnosis method is proposed.Fault data of complicated chemical process have some difficult-to-handle characteristics,such as high-dimension,non-linear and non-Gaussian distribution,so we use margin discriminant projection(MDP)algorithm to reduce dimensions and extract main features of fault data.Then,the affinity propagation(AP)clustering method is used to select core data and boundary data as the training samples to reduce memory consumption and shorten learning times.Afterwards,an integrated ENN classifier based on MultiBoost strategy is constructed to identify fault types.The artificial data sets are tested to verify the effectiveness of the proposed method and make a detailed sensitivity analysis for the key parameters.Finally,a real industrial system--Tennessee Eastman(TE)process is employed to evaluate the performance of the proposed fault diagnosis method.And the results show that the proposed method is efficient and capable to diagnose various types of faults in complicated chemical process.4.To solve the ill-posed and nonlinear inverse problem of ECT image reconstruction,a new ECT images reconstruction method based on fast linearized alternating direction method of multipliers(FLADMM)is proposed in this paper.On the basis of theoretical analysis of compressed sensing(CS),the data acquisition of ECT is regarded as a linear measurement process of permittivity distribution signal of pipe section.A new measurement matrix is designed and L1 regularization method is used to convert ECT inverse problem to a convex relaxation problem which contains prior knowledge.A new fast alternating direction method of multipliers which contained linearized idea is employed to minimize the objective function.Simulation data and experimental results indicate that compared with other methods,the quality and speed of reconstructed images is markedly improved.Also,the dynamic experimental results indicate that the proposed algorithm can fulfill the real-time requirement of ECT systems in the application.