Research On The Method Of Multi-sensor Fault Diagnosis For Truck Scale

As a weighing apparatus for large quantities of goods,a truck scale is widely used in mines,ports,storage and transportation departments.Whether it works normally or not is directly related to the weighing accuracy.The existing fault detection methods for truck scales mostly rely on manual recording and judgment,which has low detection efficiency and complicated process.Aiming at the shortcomings of the existing fault detection methods for truck scales,with the support of the National Natural Science Foundation of China“NewNeural Network and Its Application Research in Large Scale Weighing Apparatus”,the following research is carried out:This thesis introduced the research status and development trend of truck scales,as well as fault detection methods,the structure and principle of truck scales weighing system.It expounded the existing fault detection methods for truck scales and the shortcomings,and pointed out the research background and significance.By analyzing the reasons and types of the zero fault of the truck scale,this paper proposed an on-line detection method of sensor zero fault based on recursive principal component analysis and four fault detection indexes.A recursive principal component model that can be updated in real time was established,and a comprehensive evaluation method based on zero faults of load cell was realized by using fault detection indexes(T~2,T_H~2,SPE,PVR).Simulation experiment was used to verify the accuracy of the method.By analyzing the reasons and types of abrupt faults of truck scale weighing sensor,this paper presented a fault detection method of sensor abrupt change based on BP neural network and wavelet analysis,established the output prediction model of truck scale weighing sensor,and analyzed the error value between the actual output and the predicted output of the sensor by wavelet transform to realize the sudden change fault detection of truck scale.Simulation experiment was used to verify the effectiveness and accuracy of the method.A multi-sensor fault detection system for truck scale was designed,which is mainly composed of truck scale,conditioning circuit,data acquisition card and upper computer platform.The data acquisition card transmitted it to the upper computer,which realizes weighing data recording and storage,sensor zero fault detection,BP neural network training and sensor abrupt fault detection,etc.The system is convenient to operate and has high feasibility.By using the truck scale experiment platform to carried out field tests,the experiment results show that when there is only one faulty weighing sensor and the fault strength is minor,the global fault alarm?3%.The algorithm combining BPNN and wavelet analysis can accurately detect the sudden change fault of the truck scale weighing sensor and improve the detection efficiency,when there is one faulty weighing sensor and the strength of fault is minor,the global fault alarm?3%.