Abstract: Electrical Capacitance Tomography (ECT) is a process tomography technique based on capacitance-sensitive field, which can achieve the flow pattern detection of filled pipes by measuring the required capacitance data with sensors to meet the controllability of pipe flow patterns. The traditional ECT flow pattern recognition method has the disadavntages of low recognition rate, more complicated operation and error in imaging effect. In order to slove the problems of traditional methods, this paper proposes a flow pattern recognition algorithm based on PSO (Particle Swarm Optimization) optimized DBN-ELM (Deep Belief Networks-Extreme Learning Machine), in which the capacitance data acquisition module collects data and creates the corresponding data set, extracts the capacitance data features through DBN network, and adds ELM at the top layer of DBN identify the abstract capacitive data to complete the number of neurons in the hidden layer of DBN affects the learning ability and information processing ability of the whole model, so the PSO optimization algorithm is introduced to calculate the optimal number of neurons for each layer of Restricted Boltzmann Machine (RBM) first, which takes less time and has better imaging effect compared with other flow pattern recognition algorithms. It is of vital importance to accelerate the development of industrial intelligence.