Research On Fault Identification And Location Of Feeding Pipeline Blockage Based On CEEMD

In recent years,with the rapid adoption of advanced automated breeding technology,the diluted feeding system has become increasingly popular in domestic pig farming.Due to strict requirements for the proportion of diluted feed ingredients,when the raw materials are altered,the system may become prone to blockages,which will not only greatly limit the utilization of local materials in the breeding process,but also affect the normal operation of the feeding system.In severe cases,the feeding pipe might explode,leading to significant economic losses.It is of great significance to detect and locate pipeline blockages promptly and efficiently.At present,domestic research on pipeline faults mainly focuses on the defects of the pipeline itself,such as pipeline leakage,pipeline sag deformation,etc.,while there is relatively little research on pipeline blockage.In this study,an acoustic active detection method is employed,using Complete Ensemble Empirical Mode Decomposition(CEEMD)to decompose the acoustic feedback signals collected from the feeding pipe.The Recurrent Neural Network(RNN)-Long Short-Term Memory(LSTM)algorithm model and the energy attenuation localization method are employed to address the issue of pipeline blockage fault identification and location.The specific research content of this thesis is as follows:(1)COMSOL software was utilized to design and create a simulation model of the blockage sound field within the feeding pipe,and the simulation tests were carried out on the five conditions of three-way parts,no blockage,mild blockage,moderate blockage and severe blockage inside the feeding pipe,and the blockage caused by different types of feed was simulated to verify the influence of different types of blockage on the blockage recognition.The simulation results is verified by experiments.The results show that five different working conditions can produce specific rules of acoustic feedback signals,and different types of plugs under the same working conditions have a certain regularity of acoustic feedback signals,which verifies the theoretical feasibility of detecting feeding pipe plugging state according to acoustic feedback signals.(2)A blockage state detection system based on acoustic feedback signals is developed.CEEMD is used to decompose acoustic feedback signals,and feature vectors are extracted from the decomposed Intrinsic Mode Function(IMF)component and feature selection is carried out.BP Network(Back Propagation Network)and the RNN-LSTM algorithm were used for identification.The results indicate that,following feature selection,IMF components with higher differentiation between feature vectors can be acquired,which reduces the dimension of feature set and improves the recognition efficiency.Moreover,the RNN-LSTM algorithm model can effectively identify different plugging conditions in feeding pipelines,with higher recognition accuracy and good application value in actual production.(3)The energy attenuation positioning method was used to calculate three types of blockages with varying degrees at the same position,different positions along the same pipeline length,and different pipeline lengths with blockages at the same position.The ultimate detection distance of the test method was simulated by using COMSOL software.The results show that the detection method is less affected by the plugging state,and the calculated results are basically the same as the actual position of the plugging material in the feeding pipe,the relative error is less than 2%,which can effectively realize the location of the plugging in the feeding pipe.The simulation results show that the limit detection distance decreases with the increase of diluent density in feeding pipeline,which can provide a theoretical basis for the design of feeding pipeline clogging detection device.