Analysis And Research Of QNN Based On ACWF And WWF

The prediction of the abrupt change system is a key problem in scientific research. It also is a difficult one. To carry accurately on fault diagnosis(FD) and predicting analysis for a class of nonlinear abrupt change system, one of the key problem is its abrupt change feature. The feature is mainly reflected in aspects as following:the sys-tem status'change in the inputs and outputs and the system status parameters'abrupt change in faults or accidents. For example, the voltage, current and the size of the excitation had abrupt change feature when automatic voltage regulator of power gen-eration vehicles went wrong, as did the gas pressure, the potential and emission in the coal and gas outburst system. In order to analyze the prediction of the abrupt change system better, based on the historical experience data of abrupt change system, we can study and analyze the prediction method of the system, and utilize the feature of prob-ability wave of quantum wave function to construct a predictive modeling of quantum neural networks(QNN). The modeling can generalize the regularity of revulsion by understanding and learning these data. So it can be used successfully to predict this class of system.By combining the neural computation and quantum computing, an abrupt change system by wave function can be constructed, and weight wave function(WWF) is used to approach the weight value of traditional neural networks. On this basis we propose the predictive modeling of quantum neural networks with WWF. The neural networks use WWF to evolution the weight of QNN. The WWF is characterized by nonlinear, parallel, entanglement and interference. By using these advantages of QNN, we can get the diagnosis and predictive accurate results for mechanical fault diagnosis, coal and gas burst-outs predictive analysis. The method is to construct the harmonic su-perposition of quantum energy level by using harmonic wave function and gradient operators. It can express the abrupt change feature of the system. The wave function can learn the random data, that composed the kernel contents of QNN. In other words, the WWF can approach the weights of QNN through gradient algorithm. From path integral calculation and analysis of deduction of the quantum equation, this WWF de-pends on the evolution of the dissemination nucleus. the evolution of the dissemination nucleus may be a random path. If we take one image path, we can calculate the wave function weights. The concrete method is:in the error function based on traditional gradient BP algorithm, we can use the WWF to approach the weight of traditional neural networks. Suppose if|Wi(m+1)}≈|Wi(m)>, we can obtain the WWF from the error function, thus we can bridge the conversion of data space, that is to transform the neural network's computation in the plane space to the Hilbert space. It makes full use of the parallelism, probabilistic of quantum computing and the non-linearity of the neural network. So it also completes the complex mapping study of the abrupt change data.The fault diagnosis or bursts-out prediction methods are realized through the un-dulation interference of the wave function. According to the basic view spots of quan-tum information, we can construct abrupt change system with quantum superimposi-tion condition based on the wave function. The wave function revolves the informa-tion under the effect of the operator and collapses the evolved abrupt change system. It realized the quantum decoherence role of the quantum mechanism in the non-linear quantum neural units. Further, the evolved operator functions in W-QNN, motivates useful information in the network due to the coherent interaction between the QNN and environment, and extracts the network information by measuring and information collapsing, to complete the forecast analysis of the model data. In brief, QNN model of the WWF is an advanced non-linear predictive network. It not only can realize the mapping ability of predicting data, but also can realize the predictive ability of mapping data. This article will use the following idea to analyze and research the prediction of the abrupt change system. First, we will establish the probability com-putation of the abrupt change wave function(ACWF). Secondly, we will analyze the evolved computation of dissemination nucleus and the WWF. Finally, we will apply two examples:one is the mechanical electronic FD, and the other is the prediction of the coal and gas bursts out. The simulation results of the two examples showed that this method is effective and practical.