Fracturing Truck Intelligent Fault Diagnosis Method Based On Vibration Characteristics

A fracturing truck is a mechanical device used to fracture a reservoir to increase production.Fracturing trucks work under very high working conditions such as high temperature,high pressure,heavy load,etc.,which can easily lead to failure of the fracturing vehicle.Once the fracturing truck fails,it will cause unpredictable losses and even casualties.Therefore,it is necessary to diagnose the fracturing truck.At present,the research on fault diagnosis of fracturing vehicles is relatively rare.Because the vibration signal of the fracturing truck is non-linear,the working environment causes the collected signal to contain a lot of noise.In addition,the structure of the fracturing truck is complicated and the excitation source is numerous,which makes it difficult to identify the fault signal.For these reasons,it is difficult to diagnose the fracturing vehicle.Aiming at some problems such as complex vibration signal,inconspicuous features and difficult diagnosis of fracturing vehicles,this paper studies the intelligent diagnosis method of the fracturing trucks.The main contents include the following four aspects:(1)The structure and working principle of the fracturing truck and its vehicle are introduced.The typical fault modes and diagnostic methods of the fracturing vehicle are analyzed.(2)Aiming at the problem that the vibration signal of the fracturing truck is complex and contains many noise signals,a method based VMD-ICA for noise reduction at the fracturing vehicle is proposed.Combining the two to perform noise reduction processing is better than a single VMD or ICA noise reduction method.The result proves that the method can effectively remove the noise signal in the vibration signal collected by the fracturing vehicle.(3)Aiming at the problem that the feature extraction of the fracturing car is difficult and the fault characteristics are not obvious,a feature extraction method based on the improved stack sparse noise reduction auto-encoder is proposed.Based on the sparse noise reduction auto-encoder,the method performs edge processing on the loss function and then stacks it.Based on this,an algorithm model is established to extract the characteristics of the vibration signal of the fracturing vehicle.The extracted features characterize the signal characteristics under different fault conditions.The effect of feature extraction is ideal.(4)Aiming at the problem that the fault of the power end of the fracturing car is difficult to identify and the diagnosis is not good,a fault diagnosis method based on the improved convolution-limited Boltzmann machine is proposed.Through the zero-padding of the input data and the introduction of the cross-entropy sparse penalty factor to the hidden unit,thefault diagnosis accuracy is improved.The experiment proves the feasibility of the method.