48 results about "Sparse methods" patented technology

The invention relates to a sparse LMS method combining zero attraction penalty and attraction compensation, belonging to the field of signal processing. The method combines zero attraction penalty and attraction compensation, divides the coefficients of the estimated filter into near zero coefficients, small coefficients and large coefficients, and then adopts different attraction methods. In each iterative update, for the near-zero coefficients of the estimated filter, only the product term in the iterative update formula is used to calculate; for the large coefficients of the estimated filter, a slight attraction compensation is performed to speed up the estimated filtering The coefficient of the filter is used to approximate the convergence speed of the large coefficient of the channel; for the small coefficient of the estimated filter, if the coefficient approximates the zero coefficient value of the channel or the large coefficient value of the channel during the iteration process, then the estimation filter is approximated as described above. method with zero and large coefficients, otherwise, a simple zero-attraction penalty is applied to the coefficient. The method has fast convergence speed, low complexity and wide range of tuning parameters.

The invention discloses a mechanical fault sparse representation method based on wolf group parameterized joint dictionary, which belongs to the technical field of mechanical fault diagnosis. The present invention combines the principle of sparse representation to propose a gray wolf optimized sparse method based on the joint dictionary, integrates the optimization algorithm into the orthogonal matching and tracking, selects the joint dictionary atom that best matches the fault through the inner product operation, and uses the optimal atom to process the fault signal Reconstruction, and then perform envelope analysis on the reconstructed signal, so as to realize the extraction of transient characteristic components in the signal, effectively characterize the fault information, and judge the fault type more accurately. The invention is suitable for fault diagnosis of rolling bearings.

The invention discloses a method for predicting the remaining life of an IGBT. By means of the phase space frame reconstruction technology, on the basis of reconstructing the phase space by differential entropy rate, the reconstructed phase space is arranged with input signal vectors of Volterra series; Considering the correlation between the input data and the target output, the optimal selection of the input data of each frame is carried out. The present invention adopts the relatively mature forward-backward algorithm and the minimum angle regression algorithm to select the better input data in the input vector to As the input of the model; on the basis of the original ELM model, add multi-response sparse regression algorithm and one-by-one extraction method to cut out useless or little hidden layer nodes, and use three mixed neuron activation functions, so that The established network is more robust and generalizable; the present invention fully considers the difference between different input pairs of prediction models, and designs an adaptive algorithm to dynamically update the prediction model of each group of input data, which greatly improves the Improved prediction accuracy.

The invention discloses a side lobe suppression method and an array sparse method for a multi-beam imaging sonar sparse array. A sparse optimization problem of an array antenna is transformed into a linear regression problem of a sparse matrix, combination learning is carried out by taking the reconstruction of multiple beam direction patterns with different pointing directions as a target task, and a sparse semicircle array model for multi-task learning is established; based on the performance requirement of a preset main side lobe, a norm regular term of 1<1 / 2> of a weighted coefficient matrix is introduced on the basis of a least squares loss function, an iterative threshold converged method and an accelerating gradient descent method are used for solving an optimal weighting coefficient, and a weighting coefficient which minimizes a side lobe peak level is solved at the same time while optimizing the position of the sparse array. According to the side lobe suppression method and the array sparse method for the multi-beam imaging sonar sparse array, the problem that the sparse solution falls into the local optimum due to the mismatching of array position and weight vector is avoided, and peak side lobe levels of multiple beams formed by the array after sparseness is effectively reduced.