8 results about "Performance recovery" patented technology

The present application relates to the technical field of system performance optimization and resource scheduling, in particular to a system performance recovery method based on equipment adjustment and a strategy support platform thereof. The method comprises: inputting system information to be recovered; sequentially performing system reorganization feasibility evaluation and optimization, function replacement and spare parts commissioning evaluation, external allocation feasibility evaluation and optimization, and generating a recovery scheme; finally integrating the scheme and encoding output by using a preset encoding rule. The equipment adjustment operation is summarized as four basic types of system reorganization, function replacement, spare parts commissioning and external allocation. The strategy support platform comprises a hardware layer and a software layer, and through the internal scheme optimization program, the method process is automatically executed, the system information is received and the encoded optimal performance recovery strategy is output. The present application realizes the automatic generation and rapid decision of the system performance recovery strategy through the structured recovery process and the unified encoding system.

The application discloses a kind of targeted performance recovery type regenerant compounding method and system, it is related to road engineering technical field, compounding method includes: obtaining the performance characterization parameter of aging asphalt in old asphalt pavement material, and constructs aging asphalt performance matrix;According to the target incorporation amount of old asphalt pavement material, the target performance matrix of regenerated asphalt is constructed;According to aging asphalt performance matrix, candidate regenerant is determined, and the performance characterization parameter of the regenerated asphalt that candidate regenerant is mixed with aging asphalt based on preset incorporation amount, and the performance matrix of regenerated asphalt is constructed;According to aging asphalt performance matrix, target performance matrix and regenerated asphalt performance matrix, multi-index mixed model and object function are constructed, and target regenerant and optimal compounding weight coefficient are determined, and the compounding scheme of targeted performance recovery type regenerant is obtained.The application can realize the recovery of aging asphalt, and is suitable for the production and application of regenerated asphalt.

This invention relates to the field of neural network computing technology and discloses a mapping method for neural network computing in a heterogeneous environment. The method includes: parsing the neural network computation graph to extract the topological convergence degree of operator nodes and the asynchronous dwell integral of tensors; monitoring the transient operational deviation rate of computing resource units to generate a performance degradation gradient; filtering operator nodes whose convergence degree and integral exceed limits, and using the inactive period formed by the synchronization time difference of predecessor branches as a timing mask to map the filtered operator nodes to computing resource units in the performance recovery period. This invention utilizes the inherent asynchronous waiting gap of the neural network topology to mask the hardware physical temperature control recovery period, eliminates the local memory backpressure caused by asynchronous tensor stacking, and makes the global performance of heterogeneous computing resources approach the sum of the theoretical physical peak values ​​of the hardware.

The application provides an entropy-aware self-distillation-based large model migration quantization performance recovery method and system, relating to the technical field of artificial intelligence. The method aims to solve the pain point of "performance collapse" of model accuracy caused by ultra-low bit quantization due to limited memory capacity and bandwidth during the migration and adaptation of large models to domestic hardware platforms. The technical path is as follows: first, determine the pre-trained large language model as the teacher model; obtain the probability distribution of the model for Chinese text processing samples in the random subset of the corpus according to the entropy-aware mechanism, and calculate the teacher and student weighting coefficients; perform parameter quantization on the pre-trained large language model to obtain the student model to be trained; input each text processing sample in the corpus into the above two models respectively to obtain the teacher and student probability distributions respectively, and fuse the two based on the two weighting coefficients to obtain the intermediate probability distribution; minimize the difference between the student model and the intermediate probability distribution through self-distillation training, continuously repair the quantization accuracy loss and realize performance recovery in backpropagation, and obtain the trained student model. Finally, apply it to the target text processing task to obtain the corresponding result. The method aims to improve the performance of ultra-low bit quantization while realizing efficient deployment and performance reconstruction of large models in the domestic environment.

The application relates to the technical field of fabrics, in particular to a composite antibacterial fiber fabric and a preparation method thereof. The preparation method comprises the following steps: pretreating and pre-adsorbing benzophenone on terylene cotton base cloth; preparing a first grafting solution containing polymerizable hydantoin monomers, N-vinyl-2-pyrrolidone and an inner salt monomer, and performing ultraviolet light-induced grafting; without water washing, the terylene cotton base cloth is directly immersed into a second grafting solution containing butyl acrylate to perform secondary grafting; and finally, the product is obtained through washing and chlorination regeneration. Through specific grafting monomers and a layered construction process, the fabric has high initial bacteriostatic rate, excellent washing resistance, multiple regeneration and performance recovery ability after resisting sweat pollution while maintaining good air permeability and mechanical properties.

This invention discloses a method for evaluating the performance recovery of an electrolysis system after a pulsing operation. The method includes the following steps: Step 1: Collecting electrolysis system performance data for a baseline period before pulsing, the pulsing period, and the recovery evaluation period after pulsing; Step 2: Preprocessing the collected data; Step 3: Based on the preprocessed data from Step 2, fitting a natural decay trend within the baseline period before pulsing and extrapolating to obtain a non-interventional prediction curve; Step 4: Calculating four evaluation indicators within the recovery evaluation period after pulsing: recovery amount, recovery stability, improvement in decay slope, and recovery sustainability. The evaluation results can be used for selecting / recommending pulsing parameter combinations to achieve performance recovery and lifespan extension under energy consumption constraints.

This invention discloses a method for evaluating the performance recovery of an electrolysis system after a pulsing operation. The method includes the following steps: Step 1: Collecting electrolysis system performance data for a baseline period before pulsing, the pulsing period, and the recovery evaluation period after pulsing; Step 2: Preprocessing the collected data; Step 3: Based on the preprocessed data from Step 2, fitting a natural decay trend within the baseline period before pulsing and extrapolating to obtain a non-interventional prediction curve; Step 4: Calculating four evaluation indicators within the recovery evaluation period after pulsing: recovery amount, recovery stability, improvement in decay slope, and recovery sustainability. The evaluation results can be used for selecting / recommending pulsing parameter combinations to achieve performance recovery and lifespan extension under energy consumption constraints.