Immune Bacteria Foraging Algorithm And Its Application On Oil Pumping Machine Fault Diagnosis

Taking problem of the fault diagnosis technique in oil extraction technology as background, bacterial foraging algorithm is improved and is applied to the parameter optimization of modified kernel least squares support vector machine, the optimized application of least squares support vector machine is applied to oil pumping machine fault diagnosis, the concrete content is as follows:First of all, three operations of BFA(bacterial foraging algorithm) algorithm are improved, and combined with clonal selection ideas in immune algorithm to form a new hybrid algorithm——immune bacteria foraging algorithm. For the chemotaxis operation, the moving step of bacteria is shorten by the iteration proceed, so that the astringency is guaranteed, and the overall searching capability of bacteria is ensured at the same time; The clonal selection ideas in immune algorithm are used to achieve bacterial cloning,high-frequency variation and random crossover of the elite group, and to guide the search algorithm to improve accuracy; The elimination and dispersal operation is improved by guaranteeing the bacterial with the highest fitting value not be dispelled to increase the astringency accuracy. The each step of the immune bacteria foraging algorithm is introduced in detail. The convergence of the immune bacteria foraging algorithm is proved, through the three benchmark function tests, and comparing with traditional BFA algorithm performance.Secondly, the measure of similarity is introduced to kernel. In this dissertation, for any new input sample, the weighted average of the known correct outputs of a number of nearest neighbors is used as the similarity metric between samples. The similarity metric is introduced into the expression of the kernel function, and the new kernel function is constructed. Proved that the improved kernel is still a Mercer kernel, In order to verify the validity of the improved kernel, the standard precision testing function of UCI database is used for SVM compare between the standard kernel and improved kernel.Finally, the least squares support vector machine of improved kernel function with immune bacterial foraging optimization algorithm is applied to fault diagnosis in oil pumping machine. Processing current curve data of pumping wells and extracting feature vector,Setting up fault symptom sets and fault set, diagnosing test samples, then compare the results of diagnosis with other diagnosis methods’ results. the diagnosis is carried out on the test sample, and the diagnosis result comparing with other diagnosis methods of diagnosis.