Multi-parameter correlation early warning method of heat exchanger based on multivariate state estimation

With the rapid development of process industry and the increasing complexity of process, heat exchangers which serve as the core equipment suffer from the difficulty of the fault early warning. This paper presents a novel method for the fault early warning of tubular heat exchangers. The monitoring indexes are constructed and selected considering the easily measured parameters including temperature, pressure and flow rate of the process fluid. The monitoring indexes are applied to represent the heat transfer and flow resistance performance of the heat exchangers. The EMA algorithm is introduced to process the data and extract the dynamic features of the indexes. A three-dimensional memory matrix is built to represent the working condition of the heat exchangers based on the theory of multiple state estimation (MSET), which avoids the influence of varying condition on the monitoring indexes. Then, the operating observation is applied to calculate the real-time deviation, which achieves the state assessment and fault early warning of the heat exchangers. The abnormal parameters are identified by calculating the real-time deviation contribution rate of each index to achieve accurate fault location.The experimental results show that the method can realize the early warning of leakage faults when the leakage volume accounts for more than 1% of the process fluid flow rate and the early warning of scaling faults when the thickness of the scale layer accounts for more than 2.2% of the pipe diameter.