Abstract The complexity of hydraulic systems of crane outriggers is growing, which demands the fault diagnosis of the systems to be faster and comprehensive。 Based on the structural characteristics of the hydraulic systems of crane outriggers, this paper proposes a rapid and extensive hardware and software architecture model of conditioning monitoring and fault diagnosis system。 In this paper, tree diagnostic method and fuzzy neutral network theory is applied; the theoretical basis as well as the implementation method for this and similar hydraulic systems' fault diagnosis is provided。72679
Keywords Cranes, Fault diagnosis, Neural network
1 Introduction
Truck crane is an important model of engineering machinery。With its growing complexity in structure and function, it is more prone to complex troubles, so that it is harder to diagnosis the fault for hydraulic system of cranes outriggers。 In such scenarios, a single theory or method, whether classic or smart, is far from sufficient to achieve fault diagnosis that are comprehensive, accurate and fast。
Nevertheless, the combination of classic method and smart, two smart methods or more, may make a good compromise between the diagnosis' accuracy and speed。 With the above understanding, this paper utilizes a combined diagnosis algorithm, which is the fuzzy neural network, for the fault diagnosis of hydraulic system of crane outriggers。 The algorithm is implemented with a hardware platform and a software model of the diagnosis that realized the condition monitoring and fault diagnosis of the hydraulic system。
2 Establishment of fault diagnosis model of the hydraulic system of outriggers based on fuzzy neural network
The fuzzy neural network (FNN) structure model collects the advantages of neural network and fuzzy theory。 The FNN used by fault diagnosis of the hydraulic system of outriggers in this paper is shown in figure 1。
The FNN is based on BP (Back Propagation) artificial neural network and uses the tandem way with fuzzy system。 The input and output of the network are fuzzy quantity and membership of some features and some models。 The network structure is pided into five layers, input layer, fuzzy layer,hidden layer, output layer, fuzzification elimination layer。
Fig。 1 Structure model of FNN algorithm
Input layer is the first layer of the network。 This layer receives input characteristic signal from outside before directly transports the characteristic signal to the second floor--fuzzy neurons。 The transfer weight is 1。 The number of nodes in the input layer depends on the number of characteristic signal of the diagnosis。
Fuzzy layer is the second layer of the network。 Its function is to calculate membership of the input characteristics signal that belongs to fuzzy set of each variable value, according to the membership functions of the fuzzy subsets。 After fuzzification, each input layer node corresponds to three fuzzy layer nodes, representing the high side, normal and the low side respectively。 Therefore, the number of nodes is 3 times of the number of input nodes。
Hidden layer is the third layer of the network。 It is used to implement the mapping from input variable fuzzy value to the output variable fuzzy value。 The activation function used is Sigmoid function。 The number of the nodes is two times of the number of fuzzy layer nodes, according to theorem of Kolmogorov。 During the training process, adjustments can be made according to different level of accuracy。 起重机液压系统支腿的智能故障诊断英文文献和中文翻译:http://www.youerw.com/fanyi/lunwen_82719.html