6 results about "Discrete variable" patented technology

This invention discloses a linear guide rail SOMC design method, equipment, medium, and program product, including: (1) constructing a simulation model and SOMC design model with static stiffness as the target and weight and maximum contact stress as constraints based on the structural characteristics and static load analysis of the linear guide rail; (2) constructing a design space, generating an initial population and establishing a database, and obtaining a set of high-influence continuous variables and a set of high-influence discrete variables through SRC and CEA sensitivity analysis methods; (3) generating corresponding candidate sets based on the two sets of variables, and obtaining a complete set of candidate guide rails through individual pairing; (4) establishing a random forest prediction model, screening excellent subsets and selecting the best offspring guide rails; (5) updating the database and prediction model after simulation evaluation, returning to step (3) until the indicators meet the standards, and outputting the optimal parameter values. This invention can effectively balance the SOMC process with stiffness as the optimization target and weight and maximum contact stress as constraints, achieving higher accuracy and better overall performance.

The disclosure provides a method and device for generating an adversarial sample, the method for generating an adversarial sample comprising: obtaining a training sample set and a machine learning model, wherein each training sample in the training sample set has a discrete field attribute, and the discrete field attribute refers to a field attribute with a discrete variable as a field value; converting the field value of each training sample on each discrete field attribute into a field value vector to obtain a vector sample; applying a perturbation to the field value vector of each vector sample on each discrete field attribute, and taking the vector sample after the perturbation as a first perturbed sample; and generating an adversarial sample based on the first perturbed sample. The method and device for generating an adversarial sample according to the disclosure solve the problem of difficulty in generating an adversarial sample for discrete variable data, and can generate an adversarial sample for adversarial learning for data with a discrete variable by converting the field value on a discrete field attribute into a field value vector.

The invention provides a switching power supply system-oriented efficient global optimal multi-physics co-simulation method and system, and the method comprises the steps: defining a hybrid design space containing discrete variables and continuous variables, and constructing a composite objective function; generating a small number of parameter combinations of the discrete variables and the continuous variables by adopting a mixed horizontal orthogonal array to form an initial sample set; performing multi-physical field simulation on the initial sample set to obtain a performance observation value; constructing an initial Gaussian process proxy model based on the initial sample set and the performance observation value; a discrete forced continuous random sampling strategy is adopted to generate a large number of global uniform samples, iteration is carried out through a Bayesian optimization framework based on the initial Gaussian process proxy model, and optimal candidate points are obtained; and when the composite objective function converges or reaches the maximum number of simulation times, outputting a global optimal design parameter combination.

The invention discloses a metamaterial structure parameter optimization method based on continuous-binary discrete mixed variables. The method comprises the following steps: firstly, establishing a metamaterial structure parameterized model which comprises an input variable, a boundary condition and a target function; establishing a kernel function of a continuous-binary discrete mixed variable, wherein the kernel function is used for identifying continuous and binary discrete variables in a Gaussian process; then, an input space is initialized based on Sobol static sampling; constructing a multi-task Gaussian process agent model, and training the agent model; and finally, constructing an acquisition function, and calculating the overall hypervolume lifting amount of the smooth expectation by using the acquisition function to guide the next sampling decision. By monitoring the average hypervolume increase amount, whether optimization needs to increase the number of input vectors or whether the input vectors are converged or not is judged. By constructing an optimization framework suitable for continuous-binary discrete mixed variables, dual-performance optimization under a small sample condition is realized, the optimal structure configuration is efficiently obtained, and the optimization precision and the convergence speed are improved.

The invention discloses an axial plunger pump residual life prediction method and system based on a physically inspired multi-stage Wiener process, and belongs to the technical field of axial plunger pump residual life prediction.The axial plunger pump residual life prediction method comprises the steps that the system analyzes a wear failure mechanism of a key friction pair of an axial plunger pump, deduces a leakage rate calculation model of each friction pair, and determines the leakage rate as a degradation index; establishing a three-stage Wiener degradation model considering the uncertainty of the change point, and reducing the influence of a change point identification error on prediction precision by introducing a fusion transition stage to replace discrete change points; a parameter estimation method fusing an EM algorithm and a Newton-Raphson algorithm is provided, the problem that the multi-stage model is high in parameter dimension and prone to falling into local optimum is solved, and the model parameter estimation precision and efficiency are improved; on the basis of the Bayesian theory, online self-adaptive updating of the hyper-parameter of the drift coefficient is achieved; and deducing a residual life probability density function in combination with a first arrival time concept. The problem of uncertainty of change points can be effectively solved, and the index precision is high.

The application relates to a data binning method, device, computer equipment and storage medium. The method comprises the following steps: acquiring business data; dividing the business data into discrete variables and continuous variables; automatically binning each continuous variable to obtain a plurality of variable binning results corresponding to the continuous variable; in response to the number of categories of the discrete variables being less than or equal to a minimum category number threshold, outputting the discrete variables; in response to there being at least one discrete variable with a category number greater than the minimum category number threshold, automatically binning each discrete variable to obtain a plurality of variable binning results corresponding to the discrete variable; according to a preset evaluation index, screening the plurality of variable binning results corresponding to each continuous variable and / or discrete variable to obtain an optimal variable binning result; and outputting the optimal variable binning result and a parameter range of a corresponding binning method. The method can reduce the system calculation pressure, and can comprehensively obtain a plurality of variable binning results of data and an optimal variable binning result.