42 results about "Sensitivity analyses" patented technology

This invention discloses a drought early warning method based on energy flux, causal relationship analysis, and machine learning. The method includes the following steps: S1, acquiring energy flux and drought indicators, determining the optimal lag time through causal reasoning, and constructing a multi-order lag feature set; S2, establishing a tree model, using regression / kernel methods and a time series model candidate set, optimizing hyperparameters through particle swarm optimization, integrating two layers in a stacked manner, adaptively optimizing the performance of comprehensive regression and event recognition, and outputting a predicted sequence; S3, setting multi-level early warning rules according to drought thresholds, mapping drought levels, and evaluating effectiveness through statistical precision and recall; S4, calculating contribution using an additive feature attribution algorithm, identifying nonlinear thresholds to form sensitivity analysis, and improving interpretability. This invention achieves a 7-11 month early warning prediction of drought based on energy flux, with a drought early warning recall rate of 66.67%-75.86%, significantly improving the accuracy and interpretability of drought early warning.

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.

This invention relates to the field of radiator structure design technology, and provides a microchannel radiator design method based on embedded irregular materials and the obtained microchannel radiator, including: constructing a general mathematical description of the topology optimization problem; obtaining input conditions and dividing the topology optimization design domain; establishing the control equations of the topology optimization model, constructing a dual objective function, and performing sensitivity analysis of the multi-objective function with respect to design variables; solving the topology optimization model, and using filtering methods and projection techniques to obtain material distribution with clear fluid-structure boundaries; obtaining the two-dimensional flow channel configuration of the radiator, performing finite element analysis to obtain the temperature field and velocity field distribution, comprehensively evaluating the heat transfer performance, and determining the two-dimensional flow channel configuration; based on the two-dimensional flow channel configuration, constructing the corresponding three-dimensional model and performing finite element analysis to obtain the temperature field and velocity field distribution, evaluating the heat transfer performance using the PEC criterion, verifying the effectiveness of the method in three dimensions, and obtaining the final topology-optimized radiator configuration.

This invention discloses a method for predicting unsteady runoff response considering forest disturbance. The method includes data decomposition of the unsteady runoff sequence to obtain a basic runoff sequence and a disturbed runoff sequence; time-varying coefficient regression to obtain a basic runoff mapping function; correlation analysis and sensitivity analysis; using the correlation analysis results to correct the sensitivity analysis results to obtain optimized disturbance sensitivity; constructing a disturbed runoff prediction model; inputting the watershed characteristics and disturbance data of the watershed to be predicted into the basic runoff mapping function and the disturbed runoff prediction model respectively to obtain predicted basic runoff and predicted disturbed runoff; and superimposing these to obtain the predicted unsteady runoff. This method can quantify the disturbed runoff resulting from the interaction of wildfire disturbance, logging disturbance, vegetation adjustment disturbance, and conventional climate disturbance, and incorporates a competition-cooperative improved double cumulative curve (C-MDMC) to quantify the interaction effects between disturbances, outputting disturbed runoff prediction results. This provides a refined decision-making basis for watershed multi-scenario adaptive water resource management.

This invention discloses a method for maintaining the service performance of machine tool feed systems based on sensitivity analysis using a twin model and a particle swarm optimization algorithm. The overall method encompasses a five-layer framework for digital twin modeling, electromechanical wear mechanism fusion and encapsulation, time-varying element sensitivity analysis, and multi-objective maintenance decision optimization. The modeling part employs an object-oriented five-layer architecture, deeply encapsulating physical mechanisms such as electromechanical coupling wear evolution based on spatial location distribution into a quasi-physical model layer, achieving accurate description and real-time updates of implicit time-varying elements such as localized guideway wear. The decision-making part uses the Sobol sensitivity analysis method to quantify the impact weight of each element on service performance, constructs a full-effect exponential-driven adaptive tightening mechanism for early warning thresholds, and combines it with a particle swarm optimization algorithm (PI-PSO) that introduces degradation inertia and physical boundary penalties to iteratively calculate the optimal maintenance scheme with the goal of minimizing maintenance costs, downtime losses, and performance degradation. This method not only solves the problem that general models are difficult to describe the underlying electromechanical-dynamic coupling mechanism but also achieves deep integration of optimization algorithms and physical degradation laws, significantly improving the service reliability and economic benefits of machine tool feed systems.

The present application provides a shale gas complex fracture network characteristic parameter intelligent inversion method and system, the method adopts embedded discrete fracture model to establish the complex fracture network shale gas fracturing productivity model of shale reservoir; after determining the parameter value space to be inverted, the basic calculation examples are obtained based on random sampling, and parameter sensitivity analysis is carried out; the intelligent agent model is used to train the machine learning model with the basic calculation examples as the training samples, and the fracturing productivity prediction model is formed; the target example parameters meeting the error condition are optimized by using the set objective function based on the calculation results of the fracturing productivity prediction model; and then the target example parameters are brought into the complex fracture network shale gas fracturing productivity model to invert the target fracture network characteristic parameter results. The scheme can overcome the problems of time and cost consumption and limited applicability of the prior art, and can realize fast and reliable fitting by combining reservoir numerical simulation, intelligent machine learning model and optimization algorithm, and improve the shale gas complex fracture network characteristic parameter inversion efficiency.

A kind of quantitative analysis method of fusing indirect economic loss of multi-element power system safety accident belongs to the technical field of power system safety risk assessment. It includes: extracting human attribute key indicators for standardization processing;Respectively constructing each quantitative model to obtain each dimension comprehensive quantitative value;Social public opinion-network fermentation-social response force-CGE fusion model is constructed, and the conduction coefficient is calibrated by ridge regression analysis method;The standardized index and quantitative value are substituted into the fusion model, and the indirect economic loss comprehensive quantitative value of the fusion social public opinion-network fermentation-social response force attribute is solved;Finally, through sensitivity analysis, the human element with high impact weight is identified, and the targeted loss prevention and control suggestion is generated. The present application fills the gap of social public opinion-network fermentation-social response force and other elements in the analysis of indirect economic loss of power system, realizes the innovative fusion of economic and sociological model, the quantitative result is accurate, and provides scientific basis for power accident loss assessment and risk prevention and control.

The application discloses a numerical control machine tool precision distribution method from geometric error to manufacturing tolerance, comprising: using small deformation hypothesis and superposition principle to obtain the influence coefficient matrix of manufacturing tolerance to position related geometric error, and combining with the machine tool motion chain to establish a manufacturing tolerance-geometric error-space error three-layer mapping model; adopting Sobol total effect sensitivity analysis of the overall workspace characteristic trajectory to obtain the total effect sensitivity index of each manufacturing tolerance; establishing a manufacturing cost evaluation model, and combining with the total effect sensitivity index to construct a cost sensitivity index; taking the minimum manufacturing cost as the target, comprehensively considering the tolerance definition domain constraint, the sorting constraint and the space error confidence constraint, and using an optimization algorithm to realize precision distribution. The application traces the precision distribution from the abstract geometric error level to the part manufacturing tolerance level, and can further realize the collaborative optimization of machine tool precision design and manufacturing cost.

This application relates to a method and apparatus for calculating the safety margin of DC tie lines based on sensitivity analysis. The method includes: acquiring grid forecast data, power plant operation plan data, and grid maintenance data corresponding to the power system; obtaining the cross-sectional sensitivity of each DC tie line based on the grid forecast data, power plant operation plan data, and grid maintenance data; obtaining the remaining transmission margin of any DC tie line based on its transmission limit and future plan data; obtaining the current restricted margin of each DC tie line relative to other DC tie lines based on the cross-sectional sensitivity of each tie line; and calculating the intersection between the restricted margin of each cross-section and the remaining transmission margin of the tie line to obtain the final safety margin. This method enables dynamic monitoring of the final safety margin of tie lines, improving the execution efficiency and safety level of grid dispatching.

This invention provides a method for optimizing minimum EEOI speed of ships based on a PI-BT network, considering ocean current uncertainties, belonging to the field of ship energy efficiency optimization and intelligent navigation technology. Based on measured ocean current data from shipping routes, this invention mines the characteristics and probability distribution of ocean current uncertainties through statistical testing and wavelet decomposition techniques; derives the mapping relationship between EEOI and main engine speed and ocean current velocity, establishing a minimum EEOI speed optimization model; identifies core sensitive parameters through sensitivity analysis; constructs a physically guided Bayesian Transformer (PI-BT) network, designing a dual-channel input embedding layer, a Bayesian Transformer encoder, an EEOI physical information constraint layer, and a multi-objective optimization output module; and constructs a multi-component total loss function to complete network training, achieving robust optimization of minimum EEOI speed of ships under ocean current uncertainties. This invention integrates temporal modeling, uncertainty quantification, and physical constraint capabilities, significantly improving the accuracy, robustness, and computational efficiency of the speed optimization scheme.

This invention relates to the field of compressor aerodynamic performance evaluation technology, and discloses a method for quantitatively evaluating the uncertainty of blade geometric deviation performance. The method includes: acquiring the original design geometry of a multi-stage compressor and performing CFD simulation to determine the target reference flow rate; extracting deviation parameters to generate a deviation geometry sample set; performing multi-row CFD simulations on the samples; dynamically adjusting the outlet back pressure to constrain each sample to converge under the target reference flow rate; acquiring aerodynamic response data; training a surrogate model using the deviation parameters and aerodynamic response data; using the surrogate model for random sampling inference to obtain the probability distribution characteristics of the aerodynamic response data; combining a sensitivity analysis algorithm to calculate the global feature importance of each geometric deviation parameter and outputting core sensitive parameters; and determining the aerodynamic performance boundary and stall risk threshold of the compressor blade under operating conditions. This invention solves the problem of simulation condition deviation, reduces the computational cost of evaluation, and provides an engineering quantitative basis for stall risk early warning.

This invention discloses a machine learning-based method and system for predicting particle migration and blockage in fractures. The method includes the following steps: First, preset data on fracture roughness coefficient (JRC), flow velocity, equivalent particle size, and particle number; construct a fracture model with realistic shape and conduct a visualization migration experiment; record fracture blockage label data; input the parameters and labels into a neural network binary classification probability model with a multilayer perceptron as its core, and obtain a blockage probability prediction function through training; based on the trained model, perform sensitivity analysis and feature importance evaluation, and output the ranking of the impact of blockage probability on JRC, flow velocity, equivalent particle size, and particle number, thereby achieving the prediction of particle blockage events within fractures. This invention studies the particle blockage mechanism inside fractures under the coupling effect of multiple factors, providing a scientific basis for the design and construction of geotechnical engineering.

This application relates to the field of turbocharger efficiency prediction technology, and discloses a turbocharger efficiency evaluation method and device based on variable structure turbine data. The method first establishes a reliable radial turbine numerical model through high-precision three-dimensional simulation and experimental verification; then, it uses sensitivity analysis and Latin hypercube sampling to screen key impeller geometric parameters and construct an efficient sample dataset; the core is the application of a neural network algorithm to establish a nonlinear prediction model with turbine reduced speed and key geometric parameters as input and turbine efficiency as output, thereby quickly and accurately obtaining the variable structure turbine efficiency characteristics under all operating conditions; finally, by dynamically identifying the real-time operating conditions of the turbocharger, and combining the turbine efficiency predicted by the neural network with the compressor efficiency calculated by actual measurements, an accurate and efficient evaluation of the overall efficiency of the variable structure turbocharger is achieved, effectively solving the problems of high prediction difficulty and low accuracy of traditional methods.

This invention relates to the field of motor dynamic performance technology, specifically disclosing a method for optimizing excitation control parameters based on a high-voltage direct-connected synchronous condenser, comprising: Step S1, defining the excitation control parameters of the high-voltage direct-connected synchronous condenser excitation system; the excitation control parameters include: the gain of the thyristor, the time constant of the integral element of the excitation regulator, the time constant of the differential element of the excitation system, and the coefficient of the thyristor regulation element; Step S2, performing sensitivity analysis based on simulation examples to obtain the influence law of each excitation control parameter on the overvoltage value; Step S3, based on the influence law, constructing a nonlinear coupled mapping model with the excitation control parameters as input and the overvoltage value as output using a neural network; Step S4, using the LMSSA sparrow search algorithm with a leader mechanism, with the minimum overvoltage value output by the mapping model as the optimization objective function, searching and optimizing the combination of excitation control parameters, and outputting the optimized combination of excitation control parameters.

This invention discloses a data-driven causal inference-based auxiliary decision-making method for low-carbon operation of power generation, comprising the following steps: (1) collecting and preprocessing data from thermal power plants; (2) constructing a causal graph reflecting the relationship between variables; (3) identifying control variables in the causal inference; (4) using a matching algorithm to calculate and filter matching values, thereby obtaining causal inference functions for intervention variables and outcome variables; and (5) evaluating the sensitivity of low-carbon operation control of thermal power plants based on the causal inference function obtained in step (4). This invention constructs a causal graph, fits the relationship between operation control quantities and carbon emissions based on data-driven causal inference, performs sensitivity analysis on carbon emission levels, and thus achieves reasonable control of carbon emissions, thereby realizing auxiliary decision-making for low-carbon operation of thermal power plants.

This invention provides a method and system for fire emergency response plan simulation and verification based on case-based reasoning. The method includes establishing a structured historical fire case database, calculating the mixed similarity between the current scenario and historical cases based on rough set theory and cloud models, and selecting similar source cases; adapting the handling strategies of the source cases to the model by constraining the model to generate an initial plan to be verified; establishing a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources, mapping the initial plan to be verified to the transition rate and initial identifier in the database within the model, performing concurrent conflict detection and temporal logic deduction, and evaluating transient performance indicators; if the target rescue success indicator does not reach a preset threshold, iteratively correcting the resource scheduling parameters of the plan based on the sensitivity analysis results of key transitions, and repeating the deduction steps until the termination condition is met, outputting the target fire emergency response plan.

This invention provides a method, apparatus, and server for evaluating the confidence of a reused model, relating to the technical field of model evaluation. The method includes: comparing a reused model with an instance library to determine the similarity of conditional parameters, and constructing a similarity matrix using the conditional parameter similarity; performing conditional parameter sensitivity analysis on the reused model to determine sensitivity weights; using a preset similarity calculation model, based on the sensitivity weights and the similarity matrix, determining the model similarity between the reused model and each model in the instance library; and using an error expectation calculation model, based on the model similarity, determining the expected error between the simulated value of the reused model and the experimental value of the actual model, and determining the confidence of the reused model and the confidence evaluation result based on the expected error value. This invention can significantly improve computational accuracy and efficiency.

This invention belongs to the field of mechanical engineering, specifically relating to a rapid method for predicting the fatigue life of a polygonal irregular-shaped boom welded structure based on a Gaussian surrogate model. The technical solution is as follows: A global finite element model of the polygonal irregular-shaped boom is established, and the critical area is analyzed and determined; a local finite element model of the welded structure in the critical area is constructed, and the stress distribution is obtained through simulation calculation; based on the simulation results and sensitivity analysis, the key structural parameters affecting fatigue life are determined; hot spot stress and hot spot stress concentration factor are calculated using extrapolation; a sample dataset of multiple irregular-shaped boom structural parameters and hot spot stress concentration factors is constructed, and a Gaussian surrogate model considering multiple structural parameters and hot spot stress concentration factors is established; finally, the fatigue life of the polygonal irregular-shaped boom structure is predicted using the S-N curve of the hot spot stress and the welded structure. The method provided by this invention can quickly and efficiently predict the fatigue life of a polygonal irregular-shaped boom welded structure.

This application discloses a grounding grid corrosion diagnosis method, device, equipment, and medium based on sensitivity analysis and dimensionality reduction, relating to the field of power equipment fault diagnosis technology. The method equates the grounding grid to a resistive network. With the goal of minimizing a nonlinear diagnostic model, an objective function is constructed based on the initial resistance amplification factor of each branch and the sum of squared residuals between the measured and calculated port voltage values. An improved Hippo optimization algorithm is used to minimize the objective function, obtaining the resistance amplification factor of each branch. The sensitivity of each branch resistance to the objective function is calculated, and it is determined whether the sensitivity value is less than a preset threshold. If it is less, the branch is marked as a healthy branch. The improved Hippo optimization algorithm is used iteratively to solve for the other branches except for the healthy branches until the convergence condition is met, obtaining the target resistance amplification factor. The grounding grid corrosion state is diagnosed based on the target resistance amplification factor. This method is suitable for grounding grid systems with high dimensionality and sparse corrosion branches, accurately locating corrosion branches and their degree of corrosion.

The present application relates to the technical field of non-variable displacement pump design, and discloses a micro centrifugal pump blade setting angle correction method based on Reynolds number sensitivity analysis, which comprises the following steps: firstly, constructing a reference parameterized geometric model and a calculation domain; secondly, based on the dynamic similarity principle, constructing an equivalent Reynolds number working condition library by adjusting the fluid dynamic viscosity, performing CFD numerical simulation and extracting a velocity vector; then, calculating an actual fluid outflow angle, constructing a Reynolds number sensitivity function and a slip amplification coefficient; finally, deriving a target theoretical outflow angle based on a target lift, combining a prediction model to perform reverse amplification compensation on the blade setting angle, and automatically reconstructing a corrected impeller entity model. The present application quantifies the nonlinear slip rule under micro scale, ensures that the micro impeller can still output the expected theoretical lift under extremely small space and extremely low Reynolds number, realizes lift standard without repeatedly manufacturing physical prototypes, and greatly reduces trial and error costs.

This invention specifically relates to a method for analyzing the structural strength sensitivity of automotive generator assemblies. The method first constructs a finite element model of the generator assembly and calculates its unit radial and torsional stiffness coefficients; then, it constructs an acoustic simulation model and obtains a dataset of the full-frequency sound transfer function of the structural surface through full-band frequency sweep simulation; next, it calculates the single-frequency noise sensitivity coefficient and combines it with the stiffness coefficient to obtain the overall noise sensitivity coefficient; finally, it completes the sensitivity analysis and performance level assessment of the structural strength's impact on noise through preset rating rules. This invention eliminates the need for complex multi-physics coupling simulations and physical prototypes, enabling fully digital analysis and risk prediction to be completed early in the design process, significantly shortening the product development cycle and providing standardized support for low-noise forward design of generators.

The application provides a process parameter optimization method for large-tow carbon fiber pre-oxidation process, and relates to the field of computer simulation. The method comprises the following steps: a fluid domain model corresponding to a pre-oxidation furnace is established through a three-dimensional modeling technology, and a plurality of fiber tows running side by side are defined as porous medium areas; a multi-physical field cooperative strong coupling model is constructed based on the fluid domain model; process parameters in the carbon fiber pre-oxidation process are acquired, and based on the process parameters and a multi-physical process state field set, a sensitivity analysis method is used to reduce the optimization dimension to obtain target key variables; an improved non-dominated sorting genetic algorithm is used to construct a plurality of optimization objectives, and a process parameter optimization scheme is generated by defining preset constraint conditions; and based on the process parameter optimization scheme, the fluid domain model is corrected online through digital twinning technology. The application solves the problem that the existing carbon fiber pre-oxidation process control system cannot accurately reflect the real temperature distribution inside the tow.

The application provides a control surface structure flutter margin optimization method based on a Kriging surrogate model, comprising the following steps: establishing a control surface structure reference model; performing quality discrete processing on the reference model, updating the quality of each point of the control surface structure reference model according to the perturbation method, performing control surface structure flutter analysis, performing sensitivity analysis on each quality point of the control surface, obtaining the positions of a plurality of quality points with the greatest influence on the quality point discrete flutter speed, and marking the plurality of quality points as flutter significant points; updating the quality data of the plurality of flutter significant points in batches, establishing a flutter speed surrogate model based on the Kriging method; testing the flutter speed surrogate model by using a test data set; and performing rudder surface flutter optimization analysis based on the control surface flutter speed surrogate model and a genetic algorithm. The technical scheme of the application is used to solve the technical problem that the traditional design method usually depends on experience, experiments or a large amount of calculation, and the design cycle is long and the cost is high.

The application relates to the technical field of electric appliance control, in particular to an electric appliance and a control system thereof, which comprises a feature extraction module, a distribution alignment module, a parameter trend modeling module, a sensitivity analysis module and an operation strategy optimization module.In the application, through accurate screening of environmental parameters and feature parameter sparse basis vector calculation, data redundancy is significantly reduced, the analysis efficiency and pertinence are improved, the statistical adjustment of differentiated time period distribution optimizes data consistency in combination with equipment performance parameters, accurate adaptability to complex working conditions is provided, dynamic feature modeling based on time series trends realizes future change prediction of environmental parameters, effectively enhances the prediction ability and regulation efficiency of the system, the sensitivity weight distribution clearly shows the influence of each feature parameter, optimizes the collaborative control among equipment, balances performance and energy consumption, and realizes dynamic optimization of multifunction linkage.

The application discloses a high-speed permanent magnet motor vibration noise optimization method based on Fourier transform result parameterization, establishes a motor finite element model and carries out multi-physical field simulation, extracts the amplitude of a key order harmonic in radial electromagnetic force as an optimization target, selects motor configuration parameters as optimization variables, constructs an XGBoost regression model in a global stage, adopts GCFO Gaussian elite mutation to improve CFO algorithm to optimize model super parameters, and then combines NSGA-II to obtain a Pareto solution set, in a local optimization stage, first, sensitivity analysis is carried out on variables to remove parameters and reduce the value range, then, an RSM proxy model is constructed and optimization is carried out by using NSGA-II, and DEO-MADM is introduced in both stages to screen a comprehensive optimal solution from a Pareto front.Compared with the prior art, the application replaces high-cost simulation by a proxy model, significantly improves optimization efficiency and engineering applicability, and is suitable for low-noise design of a high-speed permanent magnet motor.

This invention discloses a bridge construction deflection sensitivity analysis method, comprising the following steps: Step S1, establishing a historical database; Step S2, standardizing each parameter in the historical database; Step S3, based on the standardized parameter data, taking the average of all data for that parameter, and generating random numbers with a sample size of N from the parameter distribution; Step S4, integrating a hybrid learner; Step S5, establishing a three-dimensional deflection model of the main beam using the hybrid learner; Step S6, obtaining multiple deflection values ​​as sample output values ​​using the arithmetic mean median method; Step S7, calculating the conditional expectation of each parameter with respect to the sample output values; Step S8, calculating the magnitude of the influence of the parameters on the main beam deflection using the Sobol sensitivity analysis system. This invention, by analyzing the influence of each parameter on the main beam deflection, guides the strict observation and control of parameters with a significant impact during construction, thereby avoiding resource waste and reduced construction efficiency.

The present application relates to the technical field of intelligent prediction, in particular to a red tide occurrence probability intelligent prediction method based on neural network and key factor identification, comprising S1: obtaining historical red tide event data, multi-station marine environment monitoring data, red tide emergency monitoring data and station continuous hydrological and meteorological observation data of a target sea area, and processing to obtain a standardized monitoring data set; identifying a red tide prediction key area, extracting a multi-period monitoring sequence to form a key area time sequence sample set; S2: performing multi-scale correlation analysis, sensitivity analysis and causal correlation identification to obtain a key factor sorting result; extracting a target key factor affecting red tide occurrence, determining a threshold boundary range of each target key factor, and generating a key factor threshold representation set; S3: inputting a probability prediction network and outputting a red tide occurrence probability result; and generating red tide occurrence early warning information of the target sea area within a prediction period. The present application improves the accuracy and stability of red tide prediction.

The application relates to the field of data analysis, and discloses a city flood control and waterlogging treatment engineering system failure risk diagnosis method and system, which couples a city flood and waterlogging mechanism model with an XGBoost numerical simulation technology, constructs a city flood and waterlogging key information simulator, uses a Monte Carlo simulation technology to expand an envelope type scene sample library, obtains a city flood and waterlogging key information simulation sample set based on the mechanism model and the XGBoost model, comprehensively considers system failure reasons to construct a fault tree and carries out failure probability calculation based on a Bayesian network, calculates a prior probability of engineering failure, uses reverse reasoning and sensitivity analysis of the Bayesian network, and comprehensively considers flood and waterlogging influence to complete failure risk evaluation of key subsystems, so as to identify weak links of the city flood control and waterlogging treatment engineering system. The application can reasonably evaluate the overall performance of the city flood control and waterlogging treatment engineering system and the flood and waterlogging risk level of the subsystem, avoids interference of complex factors in actual application, and improves the accuracy and comprehensiveness of diagnosis.

This invention discloses an adaptive, hardware-friendly hybrid quantization method based on deep neural networks. By analyzing the sensitivity of different network layers to quantization errors, it adaptively selects the optimal quantization strategy for each layer, thereby reducing the model's storage and computational resource consumption while maximizing model accuracy. The method includes: obtaining a full-precision deep neural network model to be quantized; extracting the weight parameters and activation parameters of each network layer; performing statistical analysis on the weight parameters and activation parameters of each network layer to construct an evaluation index for measuring quantization error; performing quantization sensitivity analysis on different network layers based on the evaluation index; adaptively determining the quantization method corresponding to each network layer based on the quantization sensitivity analysis results; performing hybrid quantization processing on the weight parameters and activation parameters according to the quantization method to obtain a hybrid precision quantization model; and completing the deep neural network inference computation based on the quantized weight parameters and activation parameters.

This invention discloses a method for predicting and warning the status trend of pumping station units. The method acquires real-time multi-source operational monitoring data of the pumping station units and inputs it into a pre-trained deep learning trend prediction model to obtain the unit status trend prediction result. Based on the prediction result and monitoring data, a multi-factor coupling sensitivity matrix is ​​calculated. The diagonal elements of this matrix represent the independent influence of individual monitoring data, while the off-diagonal elements represent the coupled interactive influence between different monitoring data. This matrix is ​​then input into a preset sensitivity pattern rule base for matching and retrieval to determine the sensitivity pattern to which the current operating state belongs. When the sensitivity pattern meets the conditions, an early warning signal is generated. This invention effectively solves the problems of poor interpretability of deep prediction models and difficulty in calculating multivariate coupling effects by introducing physical causal constraints and coupling sensitivity analysis, achieving accurate early warning with causal analysis capabilities.