Technical Studies, Remote Fault Diagnosis Based On Neural Network Construction Machinery

Remote fault diagnosis system can reach the purposes of instant response, resource sharing, remote monitoring and remote diagnosis through GPRS wireless technology linking on-site automotive terminal to remote technical diagnosis center. It not only retains the merits of traditional fault diagnosis service method, but also overcomes the constraints of the geographical and time.The engineering machinery parts are influenced by their own environment, temperature, water, dust and vibration. As the core of engineering machinery, the hydraulic system owns more complex structure. If it is in fault, it will directly affect the work efficiency, even appear major accident. Real-time detection and diagnosis on hydraulic system remote fault can shorten the downtime of engineering machinery and enhance the economic performances.This paper takes HB48 concrete pump hydraulic system in a heavy machinery company as research object. With ATmega16 microcontroller as main control center and BenQ M22A GPRS module for the transmission unit, it designs a remote data collection terminal; on the basis of hydraulic system common faults modes and network mechanisms,the BP algorithm and the BP algorithm after Hopfield optimized are applied to pump hydraulic system fault diagnosis.By researching and comparing three neural network hydraulic system fault diagnosis methods which are based on BP, H-BP and PSO,this paper proposes PSO-H-BP algorithm. It uses particle swarm algorithm to optimize the weight matrix of Hopfield network firstly, then combines BP algorithm to diagnose the fault of hydraulic system.Experiments show that using data acquisition transmission terminal formed by ATmega16 MCU and BenQ M22A can achieve real-time acquisition,rapid communication and good practical performance; compared with BP and H-BP,PSO-H-BP algorithm has higher accuracy and reliability.