| This thesis introduces the domestic and international impregnated diamond bit design and the development condition of the manufacturing first. As the technical increases increasingly along with the bit design and the quick development with the computer techniques, I discover the condition already mature to use the artificial neural network(ANN) technique in the diamond bit formulation design.Immediately after, the ANN theories are introduced. Recently the neural network makes advancement in many fields, because it has the feature of high degree of concurrent calculation and nonlinear mapping and ability of fault compatibility, and it is a self-learning and adaptive system, and it is easy to associate, especially after the advance of BP algorithm in 1986. This thesis describes the basic principle, algorithm, the shortage, the design and training processes of the BP artificial neural network.The choice and foundation of the study sample is important to set up the neural network system. This thesis adoptes the view of using value A and B of rock as indications of rock drillability in the literature[2]. The measure and formula of value A and B of rock are simply introduced. The value A of rock is defined the wear of diamond in impregnated bit. It is the main design gist of diamond parameters (grade, mesh size, concentration of diamond). The value B of rock is defined the wear of matrix in impregnated bit. It is the main design gist of bit matrix formulation design.The development of BP network system is my research scope, including the tool choice, the frame and realization of network system. The tool is Matlab6.5 software. The user interface was designed by using its GUI tool. By using the correlation functions in its neural network toolbox, the system training and simulation were realized. The network was framed, trained, simulated, saved according to the BP network model. The result of network system training was saved the Word file format. In system design, the 7 parameters such as bit outer diameter, inner diameter, the value A of rock, the value B of rock, bit weight, rotary speed and pumping quantity are the input elements of network. The 7 parameters such as the diamond grade, diamond mesh size, diamond concentration, the proportion of matrix body components, assist abrasive diamond mesh size, assist abrasive diamond concentration, coating strength WC/DIA are the output elements of network. According to the design principle of the network layer number, the system adopts an hidden layer, but its value of neuron can adjust as one's pleases. So the network structure was "7-X-7" in which the first 7 was the value of input neuron, X (variable) was the value of hidden layer and the second 7 the value of output neuron. The each layer transfer functions can be chose by Popup Menu control, having the purelin linear, the tansig hyperbolic tangent sigmoid, the logsig logarithms sigmoidtransfer function. The network training function has trainlm and trainrp. With 36 groups input vectors, the hidden layer has dimension input vectors , dimension weight matrix (biases ) and dimension output vectors . Moreover, the output layer has to has input vectors , weight matrix and dimension output vectors .This system adoptes the BP network model construction, breaking the limitation of the traditional expert system for the impregnated diamond bit design. It does not need not only a lot of production rule, but also the tree-research. It adopts the concurrent calculation, distributed storage and handling mechanism. The system has not only ability of fault compatibility and robustness but also self-learning, self-organization and adapt, and self-perfect in processing. This system based on windows platform has perfect man-machine interface. It adopts the most advanced Matlab6.5 software tool. Taking full advantage of its neural network toolbox ensures the system scientific and advanced. The program process much fast with advanced algorithm and learning rules. This system breaks BP network system in the...
|
PDF Full Text Request