160 results about "Performance comparison" patented technology

The invention provides a perceptual concrete mix proportion optimization method based on interpretable machine learning, which comprises the following steps: constructing an inductive concrete compressive strength database and performing preprocessing to determine model input characteristics, constructing an interpretable machine learning model, and performing performance comparison with other mainstream machine learning models; a feature circulation method is adopted to reduce the collinearity influence, model input features are trained based on an optimal feature function, and the contribution of the model input features to a prediction result is quantified from the overall aspect and the local aspect; preliminarily drawing up a mix proportion, substituting the mix proportion into the interpretable machine learning model after the characteristic contribution quantification, and guiding the optimal design of the mix proportion of the inductive concrete mixture according to an interpretable analysis framework. According to the method, the optimal design of the inductive concrete mixture mix proportion is guided according to an interpretability analysis framework, the contribution degree of each parameter in the mix proportion to the material performance can be quantified, and technical support is provided for rapid and intelligent design of the self-sensing concrete mix proportion based on the performance.

The invention discloses a dynamic multi-step adversarial attack method based on a graph neural network minimum disturbance budget, and belongs to the technical field of multi-step attack and information security. The dynamic multi-step adversarial attack algorithm is mainly composed of four parts of multi-step disturbance budget determination, dynamic disturbance propagation of a graph convolutional network, an optimal attack path search strategy and algorithm evaluation. Firstly, a multi-step attack loss function based on a maximized target node classification error is constructed, and accurate control over a disturbance path is achieved. Secondly, a dynamic search strategy is utilized, nodes which are most easily attacked are preferentially selected in each step, disturbance budget is reasonably distributed, and the persistence and concealment of attacks are improved through gradual diffusion; and then, combining multiple disturbance constraints of an adjacent matrix and node characteristics, and optimizing an attack process from two dimensions of disturbance intensity and balance. Finally, the effectiveness and superiority of the method provided by the invention are verified by comparing an ablation experiment with the overall performance of other similar algorithms.

The invention provides a distributed formation control method driven by a recursive balance network under a communication attack, and relates to the technical field of cooperative control of a multi-agent system, and the method comprises the steps: constructing a distributed multi-agent system model, and constructing an RBN controller heterogeneous communication network architecture and a DoS attack model; the RBN controller design is optimized, the distributed RBN controller architecture is realized, the stability of the RBN controller is analyzed based on the shrinkage mapping theory, and the exponential convergence of the system state difference is proved by constructing a Lyapunov function; training the optimized RBN controller by adopting a difficulty priority confrontation training strategy, designing a multi-objective loss function, and dynamically adjusting the training weight of each difficulty level; and simulating the distributed multi-agent system optimized by the above steps, testing the robustness of the distributed multi-agent system under incremental DoS attack intensity, and performing comparative analysis of heterogeneous and isomorphic communication network configuration and comprehensive performance comparison of an RBN controller and a traditional MPC method.

The embodiment of the invention provides a compressor performance evaluation method and device, electronic equipment and a storage medium, and belongs to the technical field of refrigeration. The method comprises the following steps: constructing a thermodynamic cycle flow of the compressor under each preset condition according to a compression process of the compressor under a plurality of preset conditions; for each preset condition, based on the thermodynamic cycle flow of the compressor under the preset condition, a mathematical model of the compressor under the preset condition is constructed, and the mathematical model comprises a plurality of target performance parameters used for evaluating the performance of the compressor; fitting the mathematical model of the compressor under the preset condition with the thermophysical parameters of the refrigerant of the compressor to obtain the value of each target performance parameter of the compressor under the preset condition; and performing performance comparison evaluation on the values of the multiple target performance parameters of the compressor under each preset condition to obtain an evaluation result. According to the embodiment of the invention, the compressor performance can be quantitatively evaluated.

The invention discloses a building composite phase change material intelligent matching and optimizing method based on a large language model, and belongs to the technical field of building energy saving and intelligent material design. Basic thermophysical property data, building environment parameter data and user demand data are collected; a deep reinforcement learning technology is utilized to construct an intelligent model based on a deep Q network strategy, and generated data samples are integrated into a thermophysical property database; outputting a candidate material combination recommendation scheme by adopting a large language model; evaluating the candidate material combination recommendation scheme by using the semantic tag vector and a sorting engine to obtain a performance evaluation result; generating a performance evaluation report according to the simulation model; a multi-agent negotiation algorithm is adopted, cross-regional thermal control performance is optimized, a multi-dimensional performance comparison diagram and tuning suggestions are generated, correction information of designers is recorded, optimization is carried out, and feedback is provided in a self-adaptive mode. According to the method, the building composite phase change material combination is screened, cross-regional thermal control is optimized, adaptive schemes and suggestions are output, and the self-adaptive optimization capability of the system is improved.

The invention provides a storage battery health state assessment method fusing ultrasonic data and hybrid machine learning, and belongs to the technical field of storage battery health state assessment. The method focuses on performance comparison of an ANN model and a GWO-ANN model in storage battery SOH regression prediction, and finds that the GWO-ANN model accurately adjusts and optimizes ANN parameters by means of a grey wolf optimization algorithm, so that the storage battery health state assessment accuracy is improved. Compared with the prior art, the method has the advantages that higher prediction precision and generalization ability are shown, the R2 value is close to perfect, the error is extremely low, and the ANN model is remarkably surpassed, the HHT-GWO-ANN model is innovatively introduced, the model integrates the advantages of HHT and GWO, the calculation process is accelerated, the robustness and prediction precision of the model are enhanced, efficient and stable model construction is achieved, and the method is suitable for large-scale popularization and application. And the test performance of the method is consistent with that of the training stage, so that the practical potential of the method in storage battery SOH nondestructive testing is verified. The ultrasonic nondestructive testing method proposed by the invention shows unique advantages in SOH recognition of the storage battery, and overcomes the limitation of a traditional method.

The invention provides a performance testing method and device for a large language model and related equipment. The performance testing method comprises the steps that multiple testing platforms for testing, testing scene configuration parameters and a structured testing data set for testing are determined; on the basis of the test scene configuration parameters and the structured test data set, synchronous constraint parameters are generated and distributed to the multiple test platforms, and then a preset load generator is used for controlling the multiple test platforms to concurrently execute reasoning test tasks about the large language model; and when the test is executed, collecting multiple dimension performance indexes of each test platform in real time for data analysis, and generating a reasoning performance comparison evaluation report of the large language model in the multiple test platforms according to a data analysis result. According to the method, the cross-platform performance evaluation period is greatly shortened, and the waste of computing resources is reduced. Manual operation errors are avoided, comparability of multi-platform performance indexes is guaranteed, and more accurate and reliable multi-platform performance test results can be obtained.

Techniques are described for providing users of an IT and security operations application with the ability to enable the collection and display of playbook run statistics. Users can selectively enable the generation of playbook run statistics for individual playbooks. Once enabled for a playbook, the IT and security operations application automatically adds source code to the playbook or otherwise enables the collection of function block-level statistics during playbook executions. Users can view the statistics collected for a playbook to compare the performance of individual blocks against one another, to compare the performance of individual playbook runs against other playbook runs or against an average of all playbook runs, and so forth. The ability to obtain playbook run statistics enables users to learn how their playbooks are performing and to troubleshoot potential issues, thereby improving the performance of playbooks and the security and operation of the IT environments in which playbooks are deployed.

The invention discloses a multi-unmanned aerial vehicle cooperative task allocation and path planning method based on a genetic algorithm, and relates to an unmanned aerial vehicle path planning method. And operating a Dijkstra algorithm to accurately solve the shortest feasible path between all airports and task points, and finally generating a global path cost matrix for quick query. A complete and safe flight path that each unmanned aerial vehicle starts from an airport, sequentially accesses task points and returns is graphically displayed, a convergence curve of an algorithm, population diversity changes and performance comparison data under different task scales, unmanned aerial vehicle numbers and obstacle densities can be output, and a parameter sensitivity analysis function is supplemented. And the effectiveness, the stability and the practical value of the proposed method in a static urban environment are comprehensively verified.

The invention relates to the field of artificial intelligence large models, and particularly provides a test method and device for large model performance transverse evaluation, and the method comprises the following steps: S1, intelligent test data generation and management; s2, performing streaming performance monitoring based on a Token timestamp; s3, dynamic concurrency and flow simulation; s4, uniformly accessing and executing multiple models; s5, analyzing and visualizing a depth result; and S6, resource monitoring. Compared with the prior art, the method has the advantages that detailed performance comparison data can be efficiently and automatically generated, and a scientific basis is provided for large model selection, performance optimization and resource scheduling.

The invention provides a roadbed performance test platform under the rainfall and load coupling effect and a use method, and aims to solve the problem that a roadbed test platform in the prior art cannot perform time-space observation and performance comparison on the same or different roadbed materials under the difference of rainfall and traffic load. Comprising a roadbed simulation mechanism, an operation simulation device, a rainfall simulation mechanism and a crushing device, the operation simulation mechanism is installed on the roadbed simulation mechanism to simulate vehicle operation, the rainfall simulation mechanism is installed above the roadbed simulation mechanism to simulate natural rainfall, and the crushing device is installed above the roadbed simulation mechanism to crush a fan-shaped roadbed after a test is completed. According to the invention, the same or different roadbed materials can be subjected to time-space observation and performance comparison under the difference of rainfall and traffic load, and the roadbed materials subjected to the test can be quickly crushed for repeated test, so that the test period is greatly shortened.

The utility model discloses a water pump motor performance comparison test bench which comprises a motor control module, a pipeline module and a cooling-water machine module, the motor control module is electrically connected with the pipeline module and the cooling-water machine module, and the cooling-water machine module is in pipeline connection with the pipeline module; the pipeline module comprises a saturator, a first control valve, a second control valve, a filter screen, a connecting pipeline and a test assembly; and the saturator is provided with a water injection port, a third control valve is arranged at the water injection port, and the saturator is connected with the water inlet and the water outlet of the cooling-water machine module through pipelines. According to the scheme, the temperature is controllable, the precision is high, the reliability is high, the stability is strong, the operation is simple, the method can be used for selecting the water pump motor with the optimal performance from a plurality of candidate matched development manufacturers, and the reliability and fairness of a comparison test are ensured at the same time.

The invention discloses an automatic performance testing method and device based on an HDFS cluster and a server. According to the method, corresponding HDFS cluster built-in test tools are called according to different test targets to execute corresponding performance test items, different scene simulations are realized through configuration of different test targets, and multi-dimensional tests can be performed for different cluster configurations. The performance in multiple aspects of throughput, metadata operation capability, high concurrency and the like can be covered, so that the performance of the cluster can be tested more comprehensively; besides, the method integrates self-contained test tools of the HDFS, and performs unified test based on standardized tools and configuration, thereby facilitating performance comparison among different clusters; according to the method, the test efficiency can be improved to the maximum extent and the human intervention is reduced through the writing of the automatic script and the assistance of the intelligent algorithm, so that the automatic performance test which is uniform in standard, efficient and reusable is realized by the method aiming at the HDFS cluster.

The invention relates to the technical field of intelligent power grid monitoring, in particular to an operation fault monitoring method and system for power equipment of a transformer substation, and the method comprises the steps: analyzing the equipment operation stability of regional nodes at different moments, and then analyzing the load fluctuation similarity between the regional nodes and adjacent regional nodes; obtaining a similarity coefficient corresponding to the region node; analyzing fluctuation difference conditions at adjacent moments and load fluctuation variation amplitudes at non-adjacent moments to obtain load mutation indexes of the regional nodes; and finally, combining the similar coefficient and the load sudden change index to obtain a load sudden change interference index of the regional node, and performing operation fault analysis on the power equipment in the transformer substation. Based on the power utilization monitoring data fluctuation and the regional power supply change, the power utilization fluctuation characteristics of each regional node are dynamically analyzed, the performance comparison is performed, the external load fluctuation and the equipment internal fault characteristics are distinguished, and the identification precision of the operation fault monitoring of the power equipment of the transformer substation can be effectively improved.

The invention provides a material database construction method and system based on high-throughput experiment multi-modal data, and relates to the technical field of material database construction, and the method comprises the steps: building a standardized experiment data collection protocol; synchronously acquiring an original multi-modal experimental data set according to a standardized experimental data acquisition protocol; processing the infrared thermal imaging data, and extracting structured temperature characteristics of each catalyst channel; integrating the experimental metadata and the sensor time sequence data to generate a standardized sample-experiment-performance associated data record; and constructing a material database by adopting a three-layer logic architecture and a mixed physical storage strategy according to a standardized sample-experiment-performance associated data record and index information of an original multi-modal experiment data set. According to the method, the material database which is standard in data, accurate in feature, clear in structure and easy to manage can be constructed. The database can effectively support data retrieval, performance comparison and model construction in material research and development.

The invention relates to the technical field of artificial intelligence and network security, in particular to a method and system for improving the question answering ability of a large model for the network security field based on large model fine tuning, and the method comprises the following steps: constructing a network security field knowledge base; selecting an appropriate large model as a basic model; annotating data in the network security field; performing performance comparison under different fine adjustment methods; a fine tuning process of the Freeze method is optimized; carrying out model evaluation and optimization, and carrying out reinforcement learning; the method and the system for improving the question-answering ability of the large model in the network security field based on the large model fine tuning have the beneficial effects that the question-answering ability of the large model in the network security field can be remarkably improved through the method for improving the question-answering ability of the large model in the network security field based on the large model fine tuning; and a more accurate and comprehensive network security knowledge service is provided for the user.

The invention provides a front-end development performance analysis and construction configuration reverse optimization method. Performance timeline data, including a resource loading time sequence, main thread task decomposition and key rendering performance indexes, of a browser developer tool are automatically collected in a running environment, and all-around monitoring of page running performance is achieved; based on the performance bottleneck of correlation between data identification and construction configuration, a low-efficiency mode of relying on artificial experience analysis is broken through; the performance problem type is mapped into the specific configuration parameter through the preset rule base, the optimization configuration item is automatically generated, and the problem that general configuration lacks pertinence is solved; an output strategy is further dynamically adjusted in combination with environmental parameters, and it is ensured that configuration adapts to multi-scene requirements; and finally, performance comparison data before and after configuration change is displayed through a visual report, a closed-loop feedback mechanism is formed, the developer optimization threshold is remarkably reduced, and the cross-environment performance consistency and iteration efficiency of front-end applications are improved.

The invention relates to the field of automobile financial consumption, and provides a method, a system, equipment and a medium for intelligently and automatically recommending an automobile loan scheme. According to the method, a two-dimensional portrait is constructed by obtaining user data and fund party data, the user data is processed by feature crossing and a portrait algorithm to generate a user portrait, and the fund party data is processed by normalization and feature engineering to generate a fund party portrait. The method is advantaged in that the user and fund party portraits are matched based on the machine learning model, the matching degree is calculated, the personalized car loan scheme is generated according to matching degree sorting, interest rate, deadline and approval requirements are considered, accurate matching of the user and the fund party is realized, and recommendation success rate and user satisfaction are substantially improved. Meanwhile, the method supports dynamic adjustment of the model, a matching algorithm is optimized in real time through performance comparison of the challenger model and the champion model, efficient extension and multi-fund-party access are achieved, and the method has industrial popularization value.

Methods, apparatus, and computer readable storage medium evaluating codec performance. One method includes obtaining m anchor data points each generated based on a respective anchor encoded video bitstream; obtaining n test data points each generated based on a respective encoded test video bitstream, n being an integer; fitting the m anchor data points with an anchor curve, the anchor curve being based on an anchor polynomial, wherein the anchor polynomial is monotonic in an x-axis range; fitting the n test data points with a test curve, the anchor curve being based on a test polynomial, wherein the test polynomial is monotonic in the x-axis range; and evaluating the test codec performance based on the anchor curve and the test curve, to obtain an evaluation result.

A system and method for quantifying the impact of investor behavioral persona on investment outcomes. A Behavioral Risk Index (BRI) is calculated based on investor behavioral factors including investor type, behavioral biases, and financial literacy. The system simulates behavior-impacted investment returns by mapping exit and re-entry points of investor groups sharing a common behavior persona during historical market events. The simulated behavior-impacted investment outcomes is quantified by comparing to buy-and-hold strategy. Personalized visualizations, including the visualization of the BRI, behavioral factor breakdowns and performance comparisons, are displayed to assist financial professionals and investors in identifying behavioral risk exposure and demonstrating the value of behavioral coaching.

The invention provides an A / B link performance comparison and strategy optimization method and system based on service flow, and belongs to the field of network performance monitoring and network strategy verification. The method comprises the following steps: firstly, performing bypass collection on real service traffic to obtain original data; copying the same copy of original data, injecting the original data into the link A and the link B at the same time, and then respectively collecting service flows of the link A and the link B; establishing a session alignment model, and judging whether the service traffic of the link A and the link B belongs to homologous traffic or not; after time drift compensation is carried out on sending timestamps of the homologous flow in the two paths, visible time sequence fields of homologous events are extracted respectively, differential indexes of the homologous events of the two paths on key time sequence parameters are calculated, and different percentile values of all the differential indexes under each parameter type in a preset time window are counted; and link performance comparison and strategy verification are carried out according to the difference index and the percentile value. According to the invention, the authenticity and efficiency of performance comparison and strategy optimization of different paths are improved.

A machine learning device includes an image set acquiring unit to acquire an image set including images, and an image set selecting unit to select an image set similar to the acquired image set from a plurality of image sets different from the acquired image set. In addition, the machine learning device includes a performance comparison unit, and a preprocessing acquisition unit to select a learning model from a plurality of machine-learned learning models based on a performance comparison result by the performance comparison unit and to acquire preprocessing performed on an image set used for machine learning of the learning model selected. Furthermore, the machine learning device includes a model learning unit to perform the acquired preprocessing on the acquired image set and to cause a learning model that has not yet trained to perform machine learning using the preprocessed image set.

The invention discloses a model poisoning attack detection method in a federated learning scene. The method comprises the following steps: S1, constructing a federated learning framework comprising an anomaly detection module based on VAE-iForest and an adaptive weight updating module; s2, dividing independent identically distributed data and non-independent identically distributed data by adopting a Dirichlet method; s3, training a VAE detection model and an iForest detection model by using the root data set in a VAE-iForest-based anomaly detection module, and performing anomaly detection on model parameter data uploaded by a client through the trained model to obtain a detection result; s4, performing performance comparison according to the detection result, calculating to obtain a weighted abnormal score, punishing the weight of the malicious client by the final detection result, and increasing the probability that the benign client is selected; and S5, dynamically adjusting the weight according to the anomaly detection effect of each model, and updating the global model for the next round of training.

The invention discloses a cross-scale virtual-real synchronization method and system for unmanned system cluster experiment teaching, and the method comprises the steps: S1, generating a corresponding digital twinborn simulation scene through a system according to a template parameter according to an algorithm template selected by a user in a graphical interface; s2, timestamp alignment and delay compensation are carried out through a cross-scale virtual-real synchronization engine, the digital twin simulation scene is synchronously mapped to a semi-physical simulation node and a physical cluster, and state cross-scale synchronization and synchronization error detection among digital twin, semi-physical simulation and the physical cluster are realized; s3, executing an algorithm task in a real object cluster, and performing real-time monitoring and control by a dynamic security fence; and S4, after the experiment is finished, generating an experiment report through the automatic evaluation engine according to the preset performance indexes. According to the invention, seamless migration among three modes can be realized without changing codes, and rapid verification, performance comparison and teaching reproduction of algorithms such as formation, obstacle avoidance and collaborative SLAM can be completed.

The invention provides a ship refrigerated container cluster redundancy control method and system, and the method comprises the steps: deploying a plurality of temperature sensors in each refrigerated container, and obtaining high-reliability environment temperature information through employing a multi-source data fusion technology; monitoring the running state of the local main controller of the refrigerated container in real time, judging whether the local main controller fails or not, and sending out an alarm signal; the redundant system monitors and stores all key temperature data of the refrigerated container in real time, and performs simple preprocessing to form a historical temperature database; comparing the control efficiency of the local main / backup controller and the control efficiency of the redundant system, and deciding an output signal weight; an accumulated historical database is utilized to dynamically adjust and optimize an adaptive fuzzy PID control rule in the redundant system; and collecting and counting models of local main / standby controllers and comparison and selection results of controller states and efficiency, and uploading the comparison and selection results to the cloud to construct an equipment efficiency knowledge base, so that faults can be accurately identified in a short time, the global temperature field characterization precision is improved, and unified regulation and control of containers of multiple brands are realized.

The present invention relates to the technical field of data detection and analysis, and discloses a database performance detection method, which detects an instruction keyword file and determines whether there is a new instruction keyword; generates a test instruction based on the new instruction keyword and performs syntax judgment, and regenerates it if it fails; adds the test instruction to a pre-stored template, parses and generates initialization data; executes the test instruction under different concurrent load conditions, collects performance indicator data and compares it with historical data, and outputs performance change results. The present invention realizes the automation and efficiency of database performance testing by dynamically generating test instructions, parsing and generating data, concurrent execution and performance comparison, and effectively solves the problems of insufficient coverage and low efficiency of manual testing in the prior art.

The invention provides a signal intensity prediction model evaluation method and device for maritime wireless communication. The method provided by the invention comprises the steps that a to-be-evaluated machine learning model set and a to-be-evaluated linear model set are determined, a plurality of to-be-evaluated models are obtained, and the outputs of the plurality of to-be-evaluated models are predicted evaluation values of signal intensity; according to the parameters of each to-be-evaluated model, sensitive features which are ranked to the first n of the output association degrees of the to-be-evaluated models are determined from the first initial feature set, wherein the sensitive features comprise features of different objects associated with ship communication; measuring a parameter set of wireless communication of the target sea area, obtaining a measurement value corresponding to the sensitive feature according to the parameter set, and constructing an evaluation data set; and evaluating each to-be-evaluated model by using the evaluation data set to obtain an evaluation result. According to the method and the device for evaluating the signal strength prediction model of the maritime wireless communication, provided by the invention, the problems that evaluation indexes are inconsistent and performance comparison cannot be intuitively carried out on the prediction model can be solved.

The invention relates to a nanoindentation and reliability characterization technology thereof, in particular to a heterogeneous material welding reliability characterization method applied to a brush type sealing structure. By means of the nanoindentation technology, mechanical properties such as hardness, elasticity modulus and plastic deformation of the welded joint are obtained by analyzing indenter deformation and indentation morphology. Meanwhile, indentation testing is carried out on different positions of the brush type sealing welding joint, and the performance comparison condition of all areas of the welding joint is obtained. Compared with a traditional reliability characterization means, the device and the method have the advantages that various areas of the heterogeneous welding joint including a base material, a heat affected zone and a fusion zone are tested under the nanoscale, various material performance indexes such as hardness, elasticity modulus and plastic deformation are measured at the same time, and the device and the method have high-precision and multifunctional testing capacity and are suitable for popularization and application. And moreover, the joint cannot be damaged or other properties cannot be influenced, and compared with a room-temperature stretching result, the nanoindentation technology has relatively high reliability and repeatability, and an accurate characterization result is obtained.

The invention relates to the technical field of sewage treatment, and discloses a multi-index comprehensive evaluation method for performance comparison and selection of a sewage treatment filler, a scoring system including physical properties, chemical properties, biological properties and economic indexes is established, the physical properties include particle size, porosity and mechanical strength, and the chemical properties include adsorbability and ion exchange capacity; the biological properties comprise microbial adhesiveness and biological membrane activity, and the economical efficiency comprises filler cost and operation and maintenance cost. According to the method, multi-dimensional index fusion is adopted, physical, porosity, mechanical strength and chemical Langmuir adsorption models are integrated, biological electron microscope observation biological membrane coverage rate and economical efficiency LCCA analysis indexes are achieved, and full life cycle evaluation is achieved.

The invention relates to a fan low-efficiency unit identification control method and a medium, and the method comprises the steps: building a multi-dimensional power curve model fusing a plurality of environmental factors based on a machine learning algorithm, deploying intelligent monitoring equipment on each fan of a wind power plant, collecting key parameters in real time, and generating a fan operation monitoring data set; according to the collected monitoring data, bin-divided BIN curve fitting is carried out on the draught fan, performance expressions of the draught fan in different wind speed and wind direction intervals are analyzed, space-time performance comparative analysis is carried out in combination with historical operation data, and by comparing performance data of the draught fan in different time periods and different wind speed intervals, a low-efficiency time period and a low-efficiency reason are determined; by constructing the multi-dimensional power curve model and combining the warehouse BIN analysis and the space-time performance comparison technology, the low-efficiency unit identification accuracy rate is 100%.