Data model system and method for intelligent reasoning of limit values ​​of test objects

Abstract

The present invention belongs to the field of data processing technology, and specifically relates to a data model system and method for intelligent reasoning of the limit values ​​of a test object. The system includes: a parameter space part, a nonlinear dynamic simulation part, and a limit reasoning part; the parameter space part is used to obtain the key parameters of the test object and determine the range of variation of each parameter; the nonlinear dynamic simulation part is used to deterministically model the test object using a nonlinear equation of a Lorentz-like system based on a discretized parameter space; the limit reasoning part is used to create a multi-scale stress field to simulate the external environment; for the potential failure matrix, the upper limit value and the lower limit value of each key parameter are calculated. The present invention realizes efficient and accurate prediction of the system's limit performance.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a data model system and method for intelligent reasoning on limit values ​​of a test object. Background Art

[0002] With the development of modern industry and technology, the performance evaluation of equipment and systems under extreme conditions has become increasingly important. In many fields, such as automobile manufacturing, aerospace, electronic equipment, etc., understanding and predicting the behavior of systems under extreme conditions is essential to ensure safety, improve performance and optimize design. The data model system of intelligent reasoning of the limit values ​​of the test object came into being. Through complex mathematical models and advanced algorithms, it can effectively predict the performance limits of equipment and systems under extreme conditions. However, there are still many deficiencies in the existing technology, which need to be further improved and perfected.

[0003] In the prior art, a variety of methods are used to predict and evaluate the extreme performance of a system. Among them, the method based on physical models simulates the behavior of a device or system under different conditions by building a detailed physical model. This method usually involves a large amount of experimental data and complex mathematical formulas. For example, finite element analysis (FEA) and multibody dynamics (MBD) are widely used in the performance evaluation of mechanical structures, and stress, strain and other key parameters are calculated through numerical simulation. Although these methods are accurate, they are computationally intensive, time-consuming, and difficult to fully consider all influencing factors for complex systems. Another type of method is based on statistical and data-driven models. These methods use historical data and machine learning algorithms to build a predictive model of the system. For example, regression analysis, support vector machine (SVM) and neural network techniques are used to predict the remaining life and failure probability of equipment. This type of method relies on a large amount of high-quality data, and the accuracy and robustness of the model will be significantly reduced when the data is insufficient or of poor quality. In addition, these models are usually black box models that lack an explanation of physical meaning, which is not conducive to a deep understanding of the failure mechanism of the system.

[0004] In recent years, the development of intelligent reasoning and optimization algorithms has provided new ideas for extreme performance prediction. Based on optimization algorithms such as genetic algorithms (GA), particle swarm optimization (PSO), and differential evolution (DE), researchers are able to find near-optimal solutions in multi-objective optimization problems. These algorithms perform well in dealing with complex nonlinear problems, but there are also some problems. For example, genetic algorithms are prone to falling into local optimal solutions and have a long calculation time; particle swarm optimization is prone to premature convergence in high-dimensional space; differential evolution performs poorly when dealing with multi-peak optimization problems. Summary of the invention

[0005] In view of this, the main purpose of the present invention is to provide a data model system and method for intelligent reasoning of the limit values ​​of a test object, thereby achieving efficient and accurate prediction of the system's extreme performance.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A data model system for intelligently reasoning about the limit values ​​of a test object, the system comprising: a parameter space part, a nonlinear dynamic simulation part and a limit reasoning part; the parameter space part is used to obtain key parameters of the test object, determine the range of variation of each parameter, construct a parameter space, discretize the parameter space to obtain a discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space; the nonlinear dynamic simulation part is used to perform deterministic modeling of the test object using a nonlinear equation of a Lorentz-like system based on the discretized parameter space, and to perform non-deterministic modeling using a random function to characterize random disturbances, and to combine the results of the deterministic modeling with the results of the non-deterministic modeling to obtain a nonlinear dynamic model of the test object; the limit reasoning part is used to create a multi-scale stress field to simulate the external environment, create an initial uniform coarse grid based on the discretized parameter space, and find a potential failure matrix in the initial uniform coarse grid through multiple iterations based on the nonlinear dynamic model, each potential failure matrix corresponds to a key parameter; for the potential failure matrix, calculate the upper limit value and the lower limit value of each key parameter.

[0008] Furthermore, the parameter space part specifically includes: a key parameter determination unit, a parameter space processing unit and a parameter starting vector creation unit; the key parameter determination unit is used to determine the parameters that affect the performance of the test object. Key parameters ;in, Indicates key parameters; for each parameter Determine its range of variation ;in, is a parameter The minimum value of is its maximum value; the parameter space processing unit is used to create a Dimensional parameter space ;in, is an n-dimensional parameter vector; choose the discrete step size Discretize each parameter to get ; The discretized parameter space is expressed as ; The parameter starting vector creation unit is used to generate a starting parameter vector for each parameter in the discretized parameter space .

[0009] Furthermore, the starting parameter vector ;in, is the starting parameter vector, is the minimum vector, is the maximum value vector, is a vector consisting of n random numbers uniformly distributed in the interval [0,1]; represents the Hadamard product.

[0010] Furthermore, the nonlinear dynamic simulation part specifically includes: a deterministic unit, an uncertain unit and a combined unit; the deterministic unit is used to perform deterministic modeling on the test object using a nonlinear equation of a Lorentz-like system, and the formula is as follows:

[0011] ;

[0012] in, is the system state vector of the test object, describing The system status, is the time limit; is the system state vector elements; It’s time; , and are all parameters of a Lorentz-like system; is an additional nonlinear function used to describe the specific behavior of the test object. By pre-defined settings, when the test object has no additional specific behavior, Each value in is 0; the non-deterministic unit is used to perform non-deterministic modeling using a random function to characterize random disturbances; the combining unit is used to combine the results of the deterministic modeling and the results of the non-deterministic modeling to obtain a nonlinear dynamic model of the test object; Results for deterministic modeling.

[0013] Furthermore, the non-deterministic unit uses a random function to perform non-deterministic modeling to characterize random disturbances through the following formula:

[0014] ;

[0015] in, It is The random perturbation intensity function of each state variable is a predefined function; is the result of non-deterministic modeling; the combination unit combines the deterministic part and the non-deterministic part through the following formula to obtain a complete nonlinear dynamic model :

[0016] ;

[0017] in, yes dimensional Gaussian white noise vector, ,in Represents a Gaussian distribution with a mean of 0 and a variance of 1; is the result of deterministic modeling.

[0018] Furthermore, in the limit reasoning part, a multi-scale stress field is created to simulate the external environment through the following formula:

[0019] ;

[0020] in, It is The amplitude of an oscillation term indicates the strength of the oscillation; It is The frequency vector of the oscillation terms represents the spatial periodicity of the oscillation; It is The phase of the oscillation term indicates the initial oscillation state; It is The decay rate of the oscillation term, which indicates how fast the oscillation decays over time; is the intensity coefficient of the nonlinear term; is the first nonlinear term The exponent of a variable controls the degree of nonlinearity; is the intensity coefficient of the fractal noise term; is the smoothness parameter of the fractal noise and They are the Fourier transform and inverse transform operators respectively; is the white noise function of the system state vector; is the wave number vector, representing frequency in Fourier space; is the system state vector The first element.

[0021] Furthermore, in the limit reasoning part, an initial uniform coarse grid is created based on the discretized parameter space through the following process: Each parameter in , choose an initial coarse grid step size , which is larger than the discretization step size :

[0022] ;

[0023] in, is an integer greater than 1; the initial uniform coarse grid It is expressed as:

[0024] ;

[0025] The total number of grid points Calculated by the following formula:

[0026] ;

[0027] in, is the scaling factor between the initial coarse grid step size and the discretization step size; is a floor function; the coordinates of the grid points are expressed as:

[0028] ;

[0029] in .

[0030] Furthermore, in the limit reasoning part, the potential failure matrix in the initial uniform coarse grid is found based on the nonlinear dynamic model through the following process:

[0031] ;

[0032] in, It is The first The key parameters are time The corresponding potential failure matrix when It is +1 iteration found +1 key parameter in time The corresponding potential failure matrix when It is an iteration parameter that controls the speed of iteration; is the stress field at location and the gradient with time t, indicating the severity of stress change; represents the second-order F-norm.

[0033] Furthermore, in the limit reasoning part, based on the potential failure matrix, the Pareto front is found, and the upper limit value in the set corresponding to the Pareto front is used as the upper limit value corresponding to the parameter; the lower limit value in the set corresponding to the Pareto front is used as the lower limit value corresponding to the parameter.

[0034] Furthermore, a data model method for intelligently reasoning the limit value of a test object comprises:

[0035] Step 1: Obtain the key parameters of the test object, determine the variation range of each parameter, construct the parameter space, discretize the parameter space to obtain the discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space;

[0036] Step 2: Based on the discretized parameter space, the nonlinear equations of the Lorentz-like system are used to perform deterministic modeling on the test object, and random functions are used to perform non-deterministic modeling to characterize random disturbances. The results of the deterministic modeling and the results of the non-deterministic modeling are combined to obtain the nonlinear dynamic model of the test object;

[0037] Step 3: Create a multi-scale stress field to simulate the external environment. Based on the discretized parameter space, create an initial uniform coarse grid. Based on the nonlinear dynamic model, find the potential failure matrix in the initial uniform coarse grid through multiple iterations. Each potential failure matrix corresponds to a key parameter. For the potential failure matrix, calculate the upper limit value and lower limit value of each key parameter.

[0038] By adopting the above technical scheme, the present invention produces the following beneficial effects: the present invention effectively improves the accuracy and reliability of prediction by combining the physical model and the data-driven model. The physical model is based on the actual operation mechanism of the equipment and system, and can provide accurate physical parameters and boundary conditions, while the data-driven model uses historical data for training and prediction, and has strong generalization ability. By combining the two, the present invention can not only utilize the precise parameters provided by the physical model, but also utilize the learning ability of the data-driven model, thereby achieving higher prediction accuracy. For example, in the performance evaluation of automobile engines, the present invention can combine the physical characteristics of the engine. The present invention introduces physical constraints and explanations into the model, thereby improving the interpretability of the data-driven model. Although the traditional black box model has high prediction accuracy, it often lacks physical explanations, which is not conducive to engineers understanding the failure mechanism of the system. By combining the physical model, the present invention makes the prediction results have clear physical meanings, helping engineers to better understand and optimize the design. For example, in the performance evaluation of the gearbox, through the combination of the physical model and the data-driven model, engineers can clearly see the impact of each parameter on the system performance, thereby carrying out targeted optimization design. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of the system structure of a data model system for intelligently reasoning about limit values ​​of test subjects provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0041] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other alternative features that are equivalent or have similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0042] Example 1: Reference Figure 1 , a data model system for intelligent reasoning of the limit values ​​of the test object, the system includes: a parameter space part, a nonlinear dynamic simulation part and a limit reasoning part; the parameter space part is used to obtain the key parameters of the test object, determine the variation range of each parameter, construct the parameter space, discretize the parameter space, obtain the discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space; the nonlinear dynamic simulation part is used to use the nonlinear equations of the Lorentz system to deterministically model the test object based on the discretized parameter space, and use random functions to perform non-deterministic modeling to characterize random disturbances, combine the results of the deterministic modeling and the results of the non-deterministic modeling, and obtain the nonlinear dynamic model of the test object; the limit reasoning part is used to create a multi-scale stress field to simulate the external environment, create an initial uniform coarse grid based on the discretized parameter space, and find the potential failure matrix in the initial uniform coarse grid through multiple iterations based on the nonlinear dynamic model, each potential failure matrix corresponds to a key parameter; for the potential failure matrix, calculate the upper limit value and the lower limit value of each key parameter.

[0043] Specifically, the system needs to obtain the key parameters of the test object, which are the key factors affecting the performance and reliability of the automobile engine or gearbox. For example, parameters such as temperature, pressure, speed and torque of the automobile engine directly affect the operating status and life of the engine. Similarly, for the gearbox, parameters such as input shaft speed, output shaft speed, input torque and output torque are also crucial. After obtaining the key parameters, the system needs to determine the range of each parameter. This range can be determined based on historical data, design specifications and experimental results. For example, the engine temperature may range from -40°C to 150°C, while the pressure may range from 0 to 200bar. These ranges provide boundary conditions for the subsequent construction of the parameter space. After determining the key parameters and their ranges, the system will construct a high-dimensional parameter space. The dimension of this parameter space is equal to the number of key parameters, and each dimension represents the range of a key parameter. The construction of the parameter space is to be able to comprehensively describe the state of the test object under various parameter combinations. Next, the system will discretize the parameter space. The purpose of discretization is to divide the continuous parameter variation range into multiple discrete points so that the parameter space can be concretized and operationalized. Discretization can be achieved through grid division or other discretization techniques. For example, if the temperature range is -40°C to 150°C, it can be divided into a discrete point every 10°C, so that multiple discrete temperature points are obtained. Other parameters are also discretized in a similar way. After discretization, the system obtains a discretized parameter space, which contains all possible parameter combinations. In the discretized parameter space, each discrete point represents a specific parameter combination. In order to effectively model and analyze these parameter combinations, the system needs to initialize a corresponding starting parameter vector for each discrete point. This starting parameter vector is based on historical data or randomly generated, and it provides initial conditions for subsequent nonlinear dynamic simulations. For example, for a discrete point combination of temperature, pressure, speed and torque, the system generates an initial parameter vector, which contains the values ​​of these parameters in the initial state. Through the above steps, the parameter space part provides a structured and highly operational data foundation for the data model system of the limit value of the intelligent reasoning test object. These discretized parameter points and initialized parameter vectors will serve as inputs for subsequent nonlinear dynamic simulations, providing the necessary prerequisites for deterministic and non-deterministic modeling. The construction and discretization of the parameter space not only enables the system to process high-dimensional data, but also provides a starting point and reference for subsequent dynamic simulations and limit value reasoning through the initial vectorization of the parameters. In this way, the entire system can accurately simulate the dynamic behavior of the test object under a variety of parameter combinations, thereby inferring its limit value.In the data model system of the limit values ​​of the test object for intelligent reasoning, the parameter space part provides solid data support and initial conditions for nonlinear dynamic simulation and limit reasoning through systematic parameter acquisition, range determination, space construction and discretization processing, so that the whole system can accurately predict the performance limits of automobile engines or gearboxes in complex environments, which has important theoretical significance and application value.

[0044] Deterministic modeling is based on nonlinear equations of Lorenz-like systems. Lorenz system is a classic chaotic system used to describe the dynamic behavior of some complex systems. Its nonlinear characteristics enable it to capture the complex relationship and mutual influence between various state variables in the system. In the present invention, nonlinear equations similar to Lorenz systems are used to model the dynamic behavior of test objects such as automobile engines or gearboxes. For example, for an automobile engine, the relationship between key parameters such as temperature, pressure, speed and torque can be described by a set of nonlinear equations, which capture the law of parameter changes over time and the interaction between them. An important step in deterministic modeling is to calibrate the model parameters to ensure that the model can accurately reflect the actual physical system. These parameters are usually obtained and adjusted through experimental data and historical records. The calibration process needs to be constantly verified and adjusted to ensure the accuracy and reliability of the model.

[0045] The purpose of non-deterministic modeling is to introduce random disturbances that may exist in the system and the uncertainty of the external environment. This part uses random functions to characterize these uncertain factors, such as fluctuations in ambient temperature, changes in material properties, and operational errors. The selection and construction of random functions are the key to non-deterministic modeling, which can be achieved using white noise or other more complex random processes. By introducing these random disturbances into the deterministic model, the system can more realistically reflect the behavior of the test object under actual operating conditions. For example, when simulating the operation of the engine, random functions can be used to simulate fluctuations in the external ambient temperature to evaluate its impact on engine performance. In this way, the system can not only predict performance under standard conditions, but also predict performance changes under various random disturbances.

[0046] After completing deterministic modeling and non-deterministic modeling, the system needs to combine the results of the two. The combination can be varied, depending on the specific application scenario and requirements. For example, a weighted average method can be used to weight the deterministic results and random results according to a certain weight; or a more complex fusion algorithm, such as Bayesian fusion, Kalman filtering, etc., can be used to optimally combine the results of the two. This combination method can effectively integrate the influence of deterministic factors and random factors, thereby obtaining a more comprehensive and accurate nonlinear dynamic model. This model can not only describe the state changes of the test object under normal operating conditions, but also simulate its behavior under extreme conditions.

[0047] An important function of the nonlinear dynamic simulation part is to verify and calibrate the simulation results. By comparing with the actual experimental data, the model parameters are adjusted to ensure the accuracy and reliability of the simulation results. This process is iterative, and through continuous optimization and improvement, the nonlinear dynamic model can more accurately reflect the actual situation of the test object. The verification and calibration process includes several steps: first, obtain actual data through experiments, which are used as the reference standard of the model; then, compare the model simulation results with the actual data to find out the differences; finally, adjust the model parameters to reduce the error. This cycle continues until the model results are highly consistent with the actual data. In this way, the system can provide high-precision simulation and prediction results, providing a scientific basis for the performance evaluation and optimization of the test object. The nonlinear dynamic simulation part also involves the initial vectorization of parameters and the use of discretized parameter space. In the discretized parameter space, each discrete point represents a specific parameter combination. In order to effectively model and analyze these parameter combinations, the system needs to initialize a corresponding starting parameter vector for each discrete point. This starting parameter vector is based on historical data or randomly generated, and it provides the initial conditions for subsequent nonlinear dynamic simulation. For example, for discrete point combinations of temperature, pressure, speed, and torque, the system generates an initial parameter vector, which contains the values ​​of these parameters in the initial state. These initial parameter vectors are used as inputs to the nonlinear dynamic simulation module, and the dynamic behavior of the test object under these parameter conditions is obtained through deterministic and non-deterministic modeling. In addition, the nonlinear dynamic simulation part can deeply explore the limit performance of the test object under different conditions through multiple iterations and simulations. This includes not only performance analysis under normal operating conditions, but also performance evaluation under extreme environments and fault conditions. Through this comprehensive simulation analysis, the system can foresee the possible failure modes and limit states of the test object under various conditions. For example, by simulating the operating state of the engine under high temperature and high pressure, its possible failure mode and limit performance can be foreseen, thus providing a reference for improved design and optimization. Overall, the nonlinear dynamic simulation part provides a powerful tool for the data model system of intelligent reasoning of the limit values ​​of the test object through comprehensive deterministic modeling and non-deterministic modeling. Through precise mathematical modeling and random processes, the system can comprehensively simulate the dynamic behavior of the test object in a complex environment and accurately predict its performance limit. The innovation of this module is that it provides a more comprehensive and accurate modeling method by comprehensively considering deterministic and random factors. It not only improves the accuracy and reliability of the model, but also provides a scientific basis and technical support for the performance evaluation and optimization of the test objects.

[0048] The limit reasoning part needs to create a multi-scale stress field to simulate the impact of the external environment on the test object. The multi-scale stress field refers to the stress applied to the test object at different scales (such as microscale and macroscale). These stresses can include mechanical stress, thermal stress, chemical stress, etc. In this way, the multiple impacts of the external environment on the test object can be simulated more comprehensively, so that the model can take into account the stress distribution at different scales. The construction of this multi-scale stress field usually requires complex numerical simulation and experimental data support. Through an accurate stress distribution model, the performance changes of the test object under different environmental conditions can be more accurately predicted. After creating the multi-scale stress field, the system creates an initial uniform coarse grid based on the discretized parameter space. This coarse grid is a rough division of the entire parameter space, and each grid point represents a specific parameter combination. In this initial stage, the role of the coarse grid is to simplify the calculation process and obtain a general understanding of the entire parameter space with less calculation. The construction of the initial uniform coarse grid needs to consider the size and complexity of the parameter space and select a suitable grid density to ensure the efficiency and accuracy of the calculation. Next, the system finds the potential failure matrix in the initial uniform coarse grid through multiple iterative analyses. The potential failure matrix refers to the state in which the test object may fail under specific parameter conditions. In each iteration, the system simulates the parameter combination of each grid point based on the nonlinear dynamic model and evaluates its response under the multi-scale stress field. Through this iterative process, the search range can be gradually narrowed down to focus on those parameter combinations that may fail. The iterative process usually uses numerical optimization methods, such as gradient descent method and genetic algorithm, to improve the efficiency and accuracy of the search. Each iteration will re-divide and refine the parameter space, making the grid denser and denser, and finally find the potential failure matrix on the high-density grid. After finding the potential failure matrix, the system needs to calculate the upper and lower limit values ​​of each key parameter. The calculation of these limit values ​​is based on the parameter combination in the potential failure matrix. Through the numerical optimization method, the limit value of each key parameter under specific stress conditions is determined. These limit values ​​are the maximum and minimum parameter values ​​that the test object can withstand under extreme conditions, which directly reflects the performance limit of the test object. For example, for an automobile engine, the calculated limit values ​​of temperature, pressure, speed and torque can be used to evaluate the performance of the engine under high temperature and high pressure conditions and predict its possible failure mode. In the process of limit reasoning, the system not only considers the deterministic nonlinear dynamic model, but also combines the influence of random disturbances, making the limit value calculation more in line with the actual situation. By combining deterministic factors and random factors, the system can more comprehensively and accurately predict the performance limit of the test object.In practical applications, this extreme reasoning method can be used to design and optimize test objects, such as improving the cooling system of an engine to increase its reliability under high temperature conditions, or optimizing the lubrication system of a gearbox to reduce wear under high pressure conditions.

[0049] Embodiment 2: The parameter space part specifically includes: a key parameter determination unit, a parameter space processing unit and a parameter starting vector creation unit; the key parameter determination unit is used to determine the parameters that affect the performance of the test object. Key parameters ;in, Indicates key parameters; for each parameter Determine its range of variation ;in, is a parameter The minimum value of is its maximum value; the parameter space processing unit is used to create a Dimensional parameter space ;in, is an n-dimensional parameter vector; choose the discrete step size Discretize each parameter to get ; The discretized parameter space is expressed as ; The parameter starting vector creation unit is used to generate a starting parameter vector for each parameter in the discretized parameter space .

[0050] Specifically, the task of the key parameter determination unit is to identify and determine the key parameters that affect the performance of the test object. These parameters are directly related to the function and reliability of the test object. By conducting detailed analysis and research on the test object, the system can identify those parameters that have a significant impact on its performance. For example, for a car engine, the key parameters may include temperature, pressure, speed and torque, while for a gearbox, they may include input shaft speed, output shaft speed, input torque and output torque. After determining these key parameters, the system also needs to determine the range of variation of each parameter, that is, ,in Representation parameters The minimum value of Indicates its maximum value. The determination of these variation ranges is usually based on a large amount of historical data, experimental results and design specifications to ensure the rationality and scientificity of the parameter range. After determining the key parameters and their variation ranges, the parameter space processing unit is responsible for creating a high-dimensional parameter space. , which is defined as This high-dimensional parameter space covers all possible parameter combinations and is the basis for subsequent modeling and analysis. However, since continuous parameter space is difficult to calculate and process, the parameter space processing unit needs to discretize it for practical operation and calculation. The discretization process is achieved by selecting an appropriate discrete step size. , each parameter Discretize within its range of variation to obtain a discretized parameter set Through this discretization method, the system can transform the continuous parameter space into a discrete parameter set to form a discretized parameter space The discretized parameter space contains all possible discrete parameter combinations, making the subsequent modeling and analysis process more practical and operational. Next, the task of the parameter starting vector creation unit is to generate a starting parameter vector for each parameter in the discretized parameter space. These starting parameter vectors are usually determined based on historical data or randomly generated to ensure that each discrete point has a reasonable initial state. For each discrete point (i.e., parameter combination), the parameter starting vector creation unit generates a vector containing the initial values ​​of all parameters. , these vectors will serve as inputs for nonlinear dynamic simulation, providing the necessary initial conditions and data support. Through these systematic steps, the parameter space part provides a solid data foundation for the data model system of the intelligent reasoning limit value of the test object. The key parameter determination unit ensures that the system can identify and process the parameters that have a significant impact on the performance of the test object and determine the reasonable range of variation of these parameters. The parameter space processing unit converts the continuous parameter space into a discrete parameter set through a discretization method, making the subsequent modeling and analysis process more practical and operational. The parameter starting vector creation unit generates an initial parameter vector for each discrete point, providing the necessary initial conditions and input data for nonlinear dynamic simulation. In specific applications, the data foundation created by the parameter space part is crucial for nonlinear dynamic simulation and limit reasoning. For example, in the nonlinear dynamic simulation part, the system needs to perform multiple iterations and simulations based on these initial parameter vectors to simulate the dynamic behavior of the test object under different parameter combinations. By comprehensively considering these key parameters and their range of variation, the system can more accurately predict the performance of the test object in a complex environment. In addition, the discretized parameter space provided by the parameter space part can also improve the efficiency and accuracy of the calculation, so that nonlinear dynamic simulation and limit reasoning can obtain accurate results in a shorter time. Through the above method, the parameter space part not only provides the necessary data support and initial conditions for nonlinear dynamic simulation and limit reasoning, but also ensures that the system can accurately simulate the dynamic behavior of the test object under various parameter combinations, thereby inferring its limit value. Each step of this process has been carefully designed and optimized to ensure the rationality, accuracy and comprehensiveness of the data, thereby providing guarantees for the performance and reliability of the entire intelligent reasoning system.

[0051] Example 3: Starting parameter vector ;in, is the starting parameter vector, is the minimum vector, is the maximum value vector, is a vector consisting of n random numbers uniformly distributed in the interval [0,1]; represents the Hadamard product.

[0052] Specifically, the starting parameter vector The generation method is based on the formula ,in represents the starting parameter vector, and are the minimum and maximum vectors of key parameters respectively, Is In A vector of uniformly distributed random numbers in the interval, represents the Hadamard product, which is element-by-element multiplication.

[0053] Specifically, we first need to determine the minimum and maximum values ​​of each key parameter, which are determined by historical data, experimental results, and design specifications to form a minimum value vector and the maximum vector For example, for parameters such as temperature, pressure, speed, and torque of a car engine, the system determines the minimum and maximum ranges of these parameters based on its design and usage specifications. Next, the system generates a random vector , where each element All in The purpose of introducing these random numbers is to ensure that the initial value of each parameter is randomly distributed within the range of its minimum and maximum values, covering all possible combinations of the entire parameter space. , the system obtains a random offset for each parameter. This offset is based on the range between the minimum and maximum values, and is multiplied by the Hadamard product of the random number so that the initial value of each parameter is evenly distributed within its possible range. For example, for a key parameter The calculation formula of its initial value is , which means that the parameter The initial value will be and Finally, by adding these random offsets to the minimum vector The system generates a complete starting parameter vector . A significant advantage of this generation method is that it can randomly generate initial conditions in the entire parameter space, thus providing a rich initial state for nonlinear dynamic simulation. This not only increases the diversity and coverage of the simulation, but also ensures that the model can be simulated and analyzed under various possible parameter combinations, thereby improving the accuracy and reliability of the prediction results. In this way, the system can comprehensively evaluate the dynamic behavior of the test object under different initial conditions and foresee its performance limit under extreme conditions. The generation of the starting parameter vector is crucial for subsequent nonlinear dynamic simulation, because accurate initial conditions can ensure the reliability of the model simulation results, thereby providing a guarantee for the overall performance of the system. Using a uniformly distributed random number generation method, the system effectively avoids the deviation of the initial conditions and ensures the comprehensiveness and representativeness of the simulation results. Through multiple simulations and iterations, the system can maintain the diversity and randomness of the results, which is particularly important when evaluating the performance of the test object in a complex and uncertain environment. For example, by randomly generating initial parameter vectors and simulating them multiple times, the system can identify potential failure modes and limit performance, providing an important reference for design optimization. In practical applications, the generation of the starting parameter vector is not only to provide the initial conditions for nonlinear dynamic simulation, but also to ensure that the system can be fully analyzed and evaluated in a wide range of parameter spaces. In this way, the system can randomly distribute the initial conditions in a complex parameter space, thus providing a broad data basis for subsequent simulation and analysis. The innovation of this method lies in its simplicity and effectiveness, and its ability to provide a rich set of initial states in a wide range of parameter spaces, ensuring that the system can fully and accurately evaluate the performance limits of the test object.

[0054] Embodiment 4: The nonlinear dynamic simulation part specifically includes: a deterministic unit, an indeterminate unit and a combined unit; the deterministic unit is used to perform deterministic modeling on the test object using the nonlinear equation of a Lorentz-like system, and the formula is as follows:

[0055] ;

[0056] in, is the system state vector of the test object, describing The system status, is the time limit; is the system state vector elements; It’s time; , and are all parameters of a Lorentz-like system; is an additional nonlinear function used to describe the specific behavior of the test object. By pre-defined settings, when the test object has no additional specific behavior, Each value in is 0; the non-deterministic unit is used to perform non-deterministic modeling using a random function to characterize random disturbances; the combining unit is used to combine the results of the deterministic modeling and the results of the non-deterministic modeling to obtain a nonlinear dynamic model of the test object; Results for deterministic modeling.

[0057] Specifically, the deterministic unit uses the nonlinear equations of a quasi-Lorenz system to perform deterministic modeling on the test object. A quasi-Lorenz system is a classic chaotic system used to describe the dynamic behavior of a complex system, and its nonlinear characteristics can capture the complex relationship between system states. In the present invention, deterministic modeling is performed by formula Implementation, where is a vector describing the state of the test object system. is the discretized parameter space, is the time variable, and are the parameters of the Lorentz-like system. In addition, a series of additional nonlinear functions are included , which is used to describe the specific behavior of the test object. Through these nonlinear equations, deterministic modeling can fully describe the behavior of the test object in different states and provide a basic dynamic model. In deterministic modeling, the state vector Each component of represents the state of the test object at different times, and these states evolve through nonlinear equations of a Lorentz-like system. For example, the first part of the equation Describes the difference and rate of change between state variables, parameters The intensity of this change is controlled. Part II Further describing the interaction between state variables, the parameters The complexity of this interaction is adjusted. The product relationship between the state variables is considered, and the parameter Through these nonlinear equations, deterministic modeling can capture the complex dynamic behavior of the test object under different conditions.

[0058] The role of the non-deterministic unit is to use random functions for non-deterministic modeling to characterize random disturbances in the system and the uncertainty of the external environment. In actual operation, the state of the test object is not only affected by internal factors, but also by the external environment and random disturbances. For example, ambient temperature, changes in material properties, operating errors, etc. These random factors can be described by adding random functions, such as white noise or other random processes. Non-deterministic modeling introduces uncertainty factors into the system through these random functions, so that the model can more realistically reflect the actual operating state of the test object. The core of non-deterministic modeling is the introduction of random disturbances, which are realized by random functions, so that the model can take into account the uncertainty of the external environment and the randomness of internal operations. For example, in the modeling of automobile engines, changes in ambient temperature may have a significant impact on engine performance. By introducing random functions, the system can simulate the impact of such temperature changes on engine performance, thereby providing a more comprehensive performance prediction. After completing deterministic modeling and non-deterministic modeling, the task of the combination unit is to synthesize the results of the two to obtain a complete nonlinear dynamic model. The combination method can be varied, such as a simple weighted average or a more complex fusion algorithm, such as Bayesian fusion or Kalman filtering. By combining deterministic results and stochastic results, the system can consider the influence of both deterministic and stochastic factors at the same time, and obtain a more comprehensive and accurate model. The combination unit can provide a more comprehensive description method of the dynamic behavior of the test object by fusing the results of the deterministic model and the non-deterministic model. The deterministic model provides a prediction of the behavior of the system under ideal conditions, while the stochastic model introduces uncertainties that may be encountered in actual operation. By combining these two parts of the results, the system can more comprehensively evaluate the dynamic behavior of the test object in a complex environment. For example, when evaluating the performance of a car gearbox, the deterministic model can provide a performance prediction of the gearbox under standard conditions, while the stochastic model can simulate the impact of environmental changes, operating errors, etc. on the performance of the gearbox. By combining these two parts of the results, the system can more accurately predict the performance of the gearbox in actual operation and foresee possible failure modes and extreme performance.

[0059] Embodiment 5: The non-deterministic unit uses a random function to perform non-deterministic modeling to characterize random disturbances through the following formula:

[0060] ;

[0061] in, It is The random perturbation intensity function of each state variable is a predefined function; is the result of non-deterministic modeling; the combination unit combines the deterministic part and the non-deterministic part through the following formula to obtain a complete nonlinear dynamic model :

[0062] ;

[0063] in, yes dimensional Gaussian white noise vector, ,in Represents a Gaussian distribution with a mean of 0 and a variance of 1; is the result of deterministic modeling.

[0064] Specifically, the random perturbation is done by the matrix To describe, each element Indicates The random perturbation intensity of each state variable. These functions are pre-set according to the specific characteristics of the test object, taking into account random factors such as external environmental changes and internal operating errors. For example, in a car engine model, changes in ambient temperature, fluctuations in fuel quality, and operating errors may all affect the performance of the engine. These factors are modeled through random perturbation functions, allowing the system to more realistically reflect actual operating conditions. The diagonal form of simplifies the computational complexity, so that the disturbance of each state variable can be processed independently, which is convenient for subsequent model combination and analysis. The introduction of random disturbances enhances the robustness of the model, allowing the model to better cope with uncertainties in the external environment and internal operations. This processing method not only improves the system's ability to simulate random events, but also ensures the reliability and effectiveness of the model in practical applications. For example, when predicting the limit performance of a car engine, by considering ambient temperature fluctuations and operating errors, the system can provide a more comprehensive performance evaluation, thereby more accurately predicting the limit state of the engine under various conditions. After completing the non-deterministic modeling, the task of the combination unit is to combine the results of the deterministic modeling with the results of the non-deterministic modeling to obtain a complete nonlinear dynamic model. This process is carried out through the formula Implementation, where It is the result of deterministic modeling and describes the dynamic behavior of the system under ideal conditions; It is the result of non-deterministic modeling and describes the random perturbations of each state variable in the system; yes dimensional Gaussian white noise vector, representing Gaussian white noise with zero mean and unit covariance. This noise model is used to simulate random fluctuations and disturbances in the external environment and internal operation. Through this combined approach, the system can consider the influence of both deterministic and random factors, thereby building a more comprehensive and accurate dynamic model. The deterministic part provides a prediction of the system's behavior under ideal conditions, while the non-deterministic part introduces uncertainties that may be encountered in actual operation, such as environmental changes and operating errors. By combining the results of these two parts, the system can not only simulate the dynamic behavior under standard conditions, but also predict performance changes under various random disturbances. The innovation of this method is that by introducing Gaussian white noise, the model can be simulated under random disturbance conditions, thus being closer to the actual operating state. The introduction of Gaussian white noise makes random disturbances statistically controllable, ensuring the stability of the model and the feasibility of calculation. For example, in the modeling of automotive gearboxes, changes in ambient temperature and operating errors may have a significant impact on gearbox performance. By introducing Gaussian white noise, the system can simulate the impact of such temperature changes and operating errors on gearbox performance, thereby providing a more comprehensive performance prediction. Through this combined method, the data model system of the limit value of the intelligent reasoning test object can comprehensively evaluate the dynamic behavior and performance limit of the test object in a complex environment. The combination of deterministic model and random model enables the system to provide accurate prediction of the ideal state and cover the actual performance under various random disturbances. This comprehensive modeling method not only improves the accuracy and reliability of the prediction results, but also provides a scientific basis and technical support for the design and optimization of the test object. In general, Example 5 constructs a comprehensive nonlinear dynamic model by introducing random disturbances and combining them with deterministic models. This method can not only capture the ideal behavior of the test object under standard conditions, but also simulate the impact of various random disturbances on the system, thereby providing a more comprehensive performance prediction. In the data model system of the limit value of the intelligent reasoning test object, the application of this method significantly improves the accuracy and reliability of the model, so that the system can accurately predict the performance limit of an automobile engine or a gearbox in a complex environment, which has important theoretical significance and practical application value.

[0065] Embodiment 6: In the limit reasoning part, a multi-scale stress field is created to simulate the external environment through the following formula:

[0066] ;

[0067] in, It is The amplitude of an oscillation term indicates the strength of the oscillation; It is The frequency vector of the oscillation terms represents the spatial periodicity of the oscillation; It is The phase of the oscillation term indicates the initial oscillation state; It is The decay rate of the oscillation term, which indicates how fast the oscillation decays over time; is the intensity coefficient of the nonlinear term; is the first nonlinear term The exponent of a variable controls the degree of nonlinearity; is the intensity coefficient of the fractal noise term; is the smoothness parameter of the fractal noise and They are the Fourier transform and inverse transform operators respectively; is the white noise function of the system state vector; is the wave number vector, representing frequency in Fourier space; is the system state vector The first element.

[0068] Specifically, the first part of the formula is the oscillatory term , which is used to simulate periodic stresses. This part of the formula is designed to take into account the periodic external disturbances to the system, such as mechanical vibration and thermal cycles. The amplitude of these oscillations represents the intensity of the oscillation, the frequency vector Represents the spatial periodicity of the oscillation, the phase describes the initial state of the oscillation, while the decay rate Controls the decay rate of oscillations over time. Through these parameters, the system can accurately simulate the impact of various periodic changes in the actual environment on the test object. For example, in the performance test of an automobile engine, mechanical vibration and thermal cycles affect the working state of the engine. These factors are accurately modeled through oscillation terms and reflected in the overall stress field of the system.

[0069] The second part of the formula Describes nonlinear stresses. This section models the effects of nonlinear external stresses on the system, such as complex loads and nonlinear deformations. Intensity coefficients for nonlinear terms controls the intensity of the overall nonlinear stress, while the exponent of each variable The degree of nonlinearity is adjusted. This design enables the system to capture the complexity and diversity of nonlinear stresses in the actual environment. For example, in the test of the gearbox, the nonlinear deformation under high load conditions has a significant impact on the performance of the system. By introducing nonlinear terms, the system can accurately describe this complex stress situation, thereby improving the accuracy of performance prediction. The third part of the formula is a fractal noise term used to simulate random disturbances. This part of the formula introduces a fractal noise model through Fourier transform and inverse transform to describe random fluctuations and disturbances in the external environment. Here, is the white noise function of the system state vector, and the wave number vector Representing frequencies in Fourier space, parameters Adjusts the smoothness of the noise, and the intensity coefficient This controls the overall intensity of the fractal noise. This design enables the system to simulate various random disturbances in the external environment, such as environmental noise and operating errors, and provide a more comprehensive stress field simulation. For example, in the performance evaluation of automobile engines, random changes in ambient temperature and fluctuations in fuel quality will affect engine performance. These random factors are modeled through fractal noise terms and thus reflected in the overall stress field of the system. Combining the design of these three parts, the formula A multi-scale stress field is created to fully simulate various stress conditions in the external environment. The oscillation term captures periodic stress, the nonlinear term describes nonlinear stress, and the fractal noise term simulates random disturbances. Through this comprehensive simulation, the system can accurately evaluate the dynamic behavior and performance limits of the test object in a complex and changing external environment. For example, when predicting the ultimate performance of an automobile gearbox, by simulating mechanical vibrations, complex loads, and random disturbances in the external environment, the system can fully evaluate the performance of the gearbox under different working conditions and predict its possible failure modes and limit states. Similarly, for other types of test objects, such as electronic equipment or mechanical systems, through the simulation of this multi-scale stress field, the system can identify potential problems under high load and complex operating conditions, providing important reference for design optimization. In practical applications, this multi-scale stress field simulation is of great significance for accurately predicting the ultimate performance of the test object. By comprehensively considering periodicity, nonlinearity, and random disturbances, the system can provide a more comprehensive and accurate performance evaluation. This not only helps to identify potential failure modes and limit states, but also provides a scientific basis for the design and optimization of the test object. For example, in the performance testing of aircraft engines, by simulating various stresses in the external environment, the system can accurately predict the performance of the engine under extreme conditions, thereby improving its reliability and safety.

[0070] Example 7: In the limit reasoning part, an initial uniform coarse grid is created based on the discretized parameter space through the following process: For the discretized parameter space Each parameter in , choose an initial coarse grid step size , which is larger than the discretization step size :

[0071] ;

[0072] in, is an integer greater than 1; the initial uniform coarse grid It is expressed as:

[0073] ;

[0074] The total number of grid points Calculated by the following formula:

[0075] ;

[0076] in, is the scaling factor between the initial coarse grid step size and the discretization step size; is a floor function; the coordinates of the grid points are expressed as:

[0077] ;

[0078] in .

[0079] Specifically, the system starts with a discretized parameter space, where each parameter has a predefined range and discrete step size. To create an initial uniform coarse grid, the system selects an initial coarse grid step size of , this step length is greater than the discrete step length ,Right now ,in is an integer greater than 1. Through this setting, the system can reduce the number of calculation points in the initial stage, thereby improving the calculation efficiency. The value can control the density of the initial coarse grid, so that the system can strike a balance between computational efficiency and accuracy. When generating the initial uniform coarse grid, the system divides the value range of each parameter so that each parameter is within the range of its minimum and maximum values ​​according to the step size. Discretization is performed. The set of these discrete points constitutes a coarse grid of the entire parameter space. Through this division, the system can perform a preliminary scan of the parameter space over a larger range, thereby identifying potential key areas. The generation of this initial coarse grid not only simplifies the calculation process, but also provides a basis for subsequent fine analysis.

[0080] The total number of grid points can be calculated by the formula This formula ensures that the system can fully cover the entire parameter space while reducing the amount of calculations. In this initial stage, the system uses a coarse grid for preliminary evaluation, and by analyzing each grid point, potential extreme performance areas in the parameter space are identified. The coordinates of each grid point are expressed as

[0081] ,in is an index ranging from 0 to the maximum integer. This formula describes the specific location of the grid points in the parameter space in such a way that the values ​​of each parameter can cover its entire range. With this preliminary partitioning of the parameter space, the system is able to identify critical areas that may affect the performance limits of the test object in a relatively short time. The creation of an initial uniform coarse grid is intended to simplify the calculation process in the preliminary stage and to provide a basis for subsequent refined analysis. By choosing a larger step size, the system is able to reduce the amount of calculations for the initial evaluation. However, this coarse grid is only used for preliminary analysis and identification of potential critical areas. After identifying these areas, the system can further refine the grid within these critical areas to improve the accuracy of the analysis. In this way, the system is able to perform a comprehensive scan of the parameter space at a low computational cost in the initial stage and identify possible extreme performance areas. The system can then use finer grids in these critical areas for detailed analysis to obtain more accurate performance predictions.

[0082] In practical applications, such as in the performance evaluation of automobile engines, the initial coarse grid can help identify potential extreme performance areas under parameter combinations such as temperature, pressure, and speed. Subsequently, the system can perform more detailed analysis in these areas to obtain more accurate performance predictions. Similarly, in the performance evaluation of transmissions, the initial coarse grid can identify potential problems under high load and complex operating conditions, and the system further refines the grid in these areas for detailed analysis, thereby providing a scientific basis for design optimization. By creating an initial uniform coarse grid, the data model system for intelligent reasoning of the limit values ​​of the test object can achieve efficient parameter space evaluation in the preliminary stage. This method effectively balances computational efficiency and prediction accuracy by reasonably selecting step size and grid points, ensuring that the system can accurately predict the performance limit of the test object in a complex environment. The creation of the initial coarse grid not only lays the foundation for the subsequent fine analysis of the system, but also provides the possibility for rapid evaluation in practical applications. Through this method, the system can better identify the extreme performance of the test object under different conditions, providing a scientific basis and technical support for design optimization and performance improvement.

[0083] Specifically, the creation of a coarse grid involves the following aspects. First, the system needs to determine the minimum and maximum ranges of each parameter, which are determined based on historical data, experimental results, and design specifications. For example, in the performance evaluation of an automobile engine, the temperature parameter may range from -40°C to 150°C, while the pressure parameter may range from 0 to 200 bar. These ranges provide boundary conditions for subsequent coarse grid division. Next, the system selects an initial coarse grid step size. , this step length is the discrete step length By choosing the appropriate The value can control the density of the coarse grid, thus achieving a balance between computational efficiency and accuracy. For example, if the discrete step size of the temperature parameter is 10°C, then the coarse grid step size It may be 20°C or 30°C. Selecting a larger step size can reduce the number of calculation points, thereby improving the calculation efficiency. When generating a coarse grid, the system follows the step size. Divide the range of values ​​for each parameter. For example, for the temperature parameter, if the minimum value is -40°C and the maximum value is 150°C, the step size is is 20°C, then the value points of the temperature parameter are -40°C, -20°C, 0°C, 20°C, 40°C, 60°C, 80°C, 100°C, 120°C, and 140°C. The collection of these value points constitutes a coarse grid of the temperature parameters. The system performs similar processing on each parameter to generate a coarse grid for the entire parameter space. In this way, the system can perform a preliminary scan of the entire parameter space in a relatively short time and identify key areas that may affect the performance limits of the test object. For example, in the performance evaluation of a gearbox, the system can identify potential problems under high load and complex operating conditions through a coarse grid, thereby providing a basis for subsequent refined analysis. The total number of grid points This formula ensures that the system can fully cover the entire parameter space while reducing the amount of calculation. Specifically, the total number It is obtained by calculating the number of points for each parameter and adding up the products. For example, if the temperature parameter has 10 points and the pressure parameter has 15 points, then the total number is =10×15=150 grid points. This approach allows the system to manage the computational effort while ensuring the comprehensiveness of the initial evaluation. After the initial coarse grid is generated, the system begins to analyze each grid point to evaluate its corresponding performance limit. This process involves simulating and calculating each parameter combination to identify possible extreme performance areas. At this stage, the system uses the coarse grid for preliminary evaluation, identifying potential extreme performance areas in the parameter space by analyzing each grid point. For example, in the performance evaluation of an automotive engine, the system can use the coarse grid to identify potential extreme performance areas under parameter combinations such as temperature, pressure, and speed. After identifying potential critical areas, the system can further refine the grid within these areas to improve the accuracy of the analysis. This process involves selecting smaller step sizes and making finer divisions of critical areas to obtain more accurate performance predictions. For example, in the performance evaluation of a transmission, the system can select smaller step sizes within the identified high load areas and perform more detailed analysis of these areas to obtain more accurate performance predictions.

[0084] Through this method, the system can perform a comprehensive scan of the parameter space at a low computational cost in the initial stage and identify possible extreme performance areas. Then, the system can use finer grids in these key areas for detailed analysis to obtain more accurate performance predictions. This method not only improves the system's initial evaluation capabilities, but also provides a solid foundation for subsequent fine analysis. In practical applications, for example, in the performance test of an aeroengine, the system can perform a more detailed analysis by identifying key parameter areas through an initial coarse grid to ensure the reliability and performance optimization of the engine under various extreme conditions. This process reflects the advantages of efficient evaluation and prediction of intelligent reasoning systems in complex environments, and provides strong support for engineering design and performance improvement. In general, Example 7 provides an efficient and comprehensive parameter space evaluation method for the data model system of the limit value of the intelligent reasoning test object by creating an initial uniform coarse grid. By selecting appropriate step sizes and grid points, the system can achieve accurate performance predictions while ensuring computational efficiency. This method not only improves the system's initial evaluation capabilities, but also provides a solid foundation for subsequent fine analysis. In practical applications, the creation of this initial coarse grid can significantly improve the efficiency and accuracy of the performance evaluation of the test object, thereby providing an important reference for design optimization. For example, in the performance test of aircraft engines, by identifying the key parameter areas with the initial coarse grid, the system can conduct more detailed analysis to ensure the reliability and performance optimization of the engine under various extreme conditions. This process reflects the advantages of intelligent reasoning systems in efficient evaluation and prediction under complex environments, and provides strong support for engineering design and performance improvement. In addition, the creation of the initial coarse grid also has important practical significance. In many engineering applications, it is not feasible to fully scan the entire parameter space because of limited computing resources. By creating an initial uniform coarse grid, the system is able to fully scan the parameter space at a lower computational cost in the preliminary stage, thereby identifying possible extreme performance areas. This approach not only improves computational efficiency, but also ensures the accuracy and comprehensiveness of the prediction results. In subsequent analysis, the system can use finer grids in the identified key areas for detailed analysis to obtain more accurate performance predictions. For example, in the performance evaluation of automotive gearboxes, the initial coarse grid can identify potential problems under high load and complex operating conditions. The system further refines the grid in these areas for detailed analysis, thereby providing a scientific basis for design optimization.

[0085] Embodiment 8: In the limit reasoning part, the potential failure matrix in the initial uniform coarse grid is found based on the nonlinear dynamic model through the following process:

[0086] ;

[0087] in, It is The first The key parameters are time The corresponding potential failure matrix when It is +1 iteration found +1 key parameter in time The corresponding potential failure matrix when It is an iteration parameter that controls the speed of iteration; is the stress field at location and the gradient with time t, indicating the severity of stress change; represents the second-order F-norm.

[0088] Specifically, this formula is used to update the potential failure matrix in each iteration. By gradually refining, the system can accurately identify and analyze the potential failure areas in the parameter space. First, the system iterates on the basis of the initial uniform coarse grid, and each iteration calculates the potential failure matrix under the current parameter combination. This matrix reflects the system at a specific time. The failure state of each key parameter is determined by the iteration process. is a parameter that controls the iteration speed and determines the size of each iteration step. , can control the convergence speed of the iterative process, so that it can quickly identify potential failure areas while maintaining the stability of the calculation. The gradient of the stress field It reflects the stress changes of the system at a specific location and time point. The larger the gradient, the more drastic the stress change and the greater the impact on failure. Therefore, by calculating the gradient of the stress field, the system can identify the areas with the most drastic stress changes, and thus focus on analyzing the potential failure risks of these areas.

[0089] State Vector The Frobenius norm of It provides the overall strength information of the system state. The larger the Frobenius norm, the higher the state strength of the system under the current parameter combination, which is very important for identifying the failure behavior of the system under high stress conditions. Combining the gradient of the stress field and the norm of the state vector, the system can more comprehensively evaluate the impact of the current parameter combination on system failure. In the iterative formula, represents the total number of initial coarse grids, and represents the volume of the discretized parameter space. These two parameters are used to normalize the iteration process to ensure that each iteration is performed within a reasonable parameter range. Through this normalization process, the system can maintain the stability and accuracy of the calculation, and the iteration results will not be affected by the parameter space being too large or too small. The time derivative of the system state It reflects the rate of change of the system state over time, which is crucial for evaluating the dynamic behavior under the current parameter combination. If the time derivative is large, it means that the dynamic changes of the system under the current parameter combination are more drastic, which has a greater impact on failure. Therefore, by calculating the time derivative of the system state, the system can evaluate the impact of the current parameter combination on the dynamic behavior, thereby more accurately identifying potential failure points during the iteration process.

[0090] Parameter combination and its higher powers Used to simulate the nonlinear effects of parameters on the state of the system. Nonlinear effects are very common in actual engineering applications. For example, under high stress or complex environmental conditions, the response of the system is often not linear. By considering the nonlinear effects of parameter combinations, the system can more accurately evaluate the degree of influence of the current parameter combination on system failure. In practical applications, this iterative method has important practical significance. For example, in the performance evaluation of automobile engines, by iteratively refining the potential failure matrix, the system can identify specific failure points under parameter combinations such as temperature, pressure and speed, thereby providing more accurate performance predictions. Similarly, in the performance evaluation of gearboxes, the system can identify potential failure points under high load and complex operating conditions through iterative methods, thereby providing a scientific basis for design optimization. Through the iterative formula, the system adjusts the grid density according to the current iteration results in each iteration, and gradually refines the potential failure matrix. This iterative refinement method ensures that the system can gradually increase the accuracy on the basis of the initial coarse grid, thereby more accurately identifying potential failure points in the parameter space. After each iteration, the system adjusts the division of the parameter space according to the new failure matrix, gradually focusing on the area where failure is most likely to occur, thereby improving the accuracy of the prediction. For example, in the performance evaluation of automobile engines, by iteratively refining the potential failure matrix, the system can identify specific failure points under parameter combinations such as temperature, pressure, and speed, thereby providing more accurate performance predictions. In the initial stage, the system may use a larger grid step size to quickly scan the entire parameter space and identify potential failure areas. As the iterations proceed, the system gradually reduces the grid step size and focuses on analyzing those high-risk areas, thereby improving the precision of the analysis and the accuracy of the prediction.

[0091] In addition, in the performance evaluation under complex operating conditions, the iterative refinement method is also of great significance. Through multiple iterations, the system can gradually identify potential failure points under high load and complex operating conditions. For example, in the test of the gearbox, the system can identify the failure mode under high load and long-term operation conditions through iteration, thereby providing a scientific basis for the design optimization of the gearbox. In general, Example 8 finds the potential failure matrix in the initial uniform coarse grid by the iterative refinement method, and provides an efficient and accurate evaluation method for the data model system of the limit value of the intelligent reasoning test object. Through the nonlinear influence of the comprehensive stress field gradient, the Frobenius norm of the system state, the time derivative and the parameter combination, the system can gradually refine the potential failure area in each iteration, thereby realizing the accurate prediction of the limit value of the test object. This method not only improves the calculation efficiency, but also ensures the accuracy and reliability of the prediction results, and has important theoretical significance and practical application value in practical engineering applications.

[0092] Embodiment 9: In the limit reasoning part, based on the potential failure matrix, the Pareto front is found, and the upper limit value in the set corresponding to the Pareto front is used as the upper limit value corresponding to the parameter; the lower limit value in the set corresponding to the Pareto front is used as the lower limit value corresponding to the parameter.

[0093] Specifically, the potential failure matrix obtained from Example 8 is start, It is expressed as:

[0094] ;

[0095] in It is potential failure state vectors, including combinations of multiple parameters. In order to identify points on the Pareto frontier, the system needs to calculate all possible parameter combinations and select the optimal combination. is a The matrix of is the number of parameter combinations, is the dimension of the parameter. A multi-objective optimization algorithm, such as the non-dominated sorting genetic algorithm (NSGA-II), is introduced to find the Pareto frontier. The steps of the non-dominated sorting genetic algorithm are as follows: Initialize the population: randomly generate Parameter combinations ,in Usually greater than . Calculate the fitness value: for each parameter combination Calculate the objective function , and determine its non-dominated level. Selection: Select according to the non-dominated level Excellent individuals are selected to form the next generation population. Crossover and mutation: Perform crossover and mutation operations on the selected individuals to generate new parameter combinations. Repeat the above steps until the predetermined number of iterations or convergence conditions are reached. Through the above steps, the system can find the Pareto frontier. After determining the Pareto front, the system needs to further analyze these points to determine the limit values ​​of each parameter. Specifically, the system uses the maximum value in the set corresponding to the Pareto front as the upper limit value of the parameter, and the minimum value in the set corresponding to the Pareto front as the lower limit value of the parameter.

[0096] Embodiment 10: A data model method for intelligently reasoning about the limit value of a test object, the method comprising:

[0097] Step 1: Obtain the key parameters of the test object, determine the variation range of each parameter, construct the parameter space, discretize the parameter space to obtain the discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space;

[0098] Step 2: Based on the discretized parameter space, the nonlinear equations of the Lorentz-like system are used to perform deterministic modeling on the test object, and random functions are used to perform non-deterministic modeling to characterize random disturbances. The results of the deterministic modeling and the results of the non-deterministic modeling are combined to obtain the nonlinear dynamic model of the test object;

[0099] Step 3: Create a multi-scale stress field to simulate the external environment. Based on the discretized parameter space, create an initial uniform coarse grid. Based on the nonlinear dynamic model, find the potential failure matrix in the initial uniform coarse grid through multiple iterations. Each potential failure matrix corresponds to a key parameter. For the potential failure matrix, calculate the upper limit value and lower limit value of each key parameter.

[0100] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A data model system for intelligent reasoning of the limit value of a test object, characterized in that: The system includes: a parameter space part, a nonlinear dynamic simulation part and a limit reasoning part; the parameter space part is used to obtain key parameters of the test object, determine the variation range of each parameter, construct a parameter space, discretize the parameter space to obtain a discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space; the nonlinear dynamic simulation part is used to perform deterministic modeling of the test object using a nonlinear equation of a Lorentz-like system based on the discretized parameter space, and to perform non-deterministic modeling using a random function to characterize random disturbances, and to combine the results of the deterministic modeling and the results of the non-deterministic modeling to obtain a nonlinear dynamic model of the test object; the limit reasoning part is used to create a multi-scale stress field to simulate the external environment, create an initial uniform coarse grid based on the discretized parameter space, and find the potential failure matrix in the initial uniform coarse grid through multiple iterations based on the nonlinear dynamic model, and each potential failure matrix corresponds to a key parameter; for the potential failure matrix, calculate the upper limit value and the lower limit value of each key parameter; The test object is an automobile engine or a gearbox; when the test object is an automobile engine, the key parameters include: temperature, pressure, speed and torque; when the test object is a gearbox, the key parameters include: input shaft speed, output shaft speed, input torque and output torque; The parameter space part specifically includes: a key parameter determination unit, a parameter space processing unit and a parameter starting vector creation unit; the key parameter determination unit is used to determine the parameters that affect the performance of the test object. Key parameters ;in, Indicates key parameters; for each parameter Determine its range of variation ;in, is a parameter The minimum value of is its maximum value; the parameter space processing unit is used to create a Dimensional parameter space ;in, is an n-dimensional parameter vector; choose the discrete step size Discretize each parameter to get ; The discretized parameter space is expressed as ; The parameter starting vector creation unit is used to generate a starting parameter vector for each parameter in the discretized parameter space .

2. The data model system for the limit value of the intelligent reasoning test object according to claim 1, characterized in that: Starting parameter vector ;in, is the starting parameter vector, is the minimum vector, is the maximum value vector, is a vector consisting of n random numbers uniformly distributed in the interval [0,1]; represents the Hadamard product.

3. The data model system for the limit value of the intelligent reasoning test object according to claim 2, characterized in that: The nonlinear dynamic simulation part specifically includes: deterministic units, non-deterministic units and combined units; the deterministic units are used to perform deterministic modeling on the test object using the nonlinear equations of the Lorentz-like system, and the formula is as follows: ； in, is the system state vector of the test object, describing The system status, is the time limit; is the system state vector elements; It’s time; , and are all parameters of a Lorentz-like system; is an additional nonlinear function used to describe the specific behavior of the test object. By pre-defined settings, when the test object has no additional specific behavior, Each value in is 0; the non-deterministic unit is used to perform non-deterministic modeling using a random function to characterize random disturbances; the combining unit is used to combine the results of the deterministic modeling and the results of the non-deterministic modeling to obtain a nonlinear dynamic model of the test object; Results for deterministic modeling.

4. The data model system for the limit value of the intelligent reasoning test object according to claim 3, characterized in that: The non-deterministic unit uses a random function to perform non-deterministic modeling to characterize random disturbances through the following formula: ； in, It is The random perturbation intensity function of each state variable is a predefined function; is the result of non-deterministic modeling; the combination unit combines the deterministic part and the non-deterministic part through the following formula to obtain a complete nonlinear dynamic model : ； in, yes dimensional Gaussian white noise vector, ,in Represents a Gaussian distribution with a mean of 0 and a variance of 1; is the result of deterministic modeling.

5. The data model system for the limit value of the intelligent reasoning test object according to claim 4, characterized in that: In the limit reasoning part, a multi-scale stress field is created to simulate the external environment through the following formula: ； in, It is The amplitude of an oscillation term indicates the strength of the oscillation; It is The frequency vector of the oscillation terms represents the spatial periodicity of the oscillation; It is The phase of the oscillation term indicates the initial oscillation state; It is The decay rate of the oscillation term, which indicates how fast the oscillation decays over time; is the intensity coefficient of the nonlinear term; is the first nonlinear term The exponent of a variable controls the degree of nonlinearity; is the intensity coefficient of the fractal noise term; is the smoothness parameter of the fractal noise and They are the Fourier transform and inverse transform operators respectively; is the white noise function of the system state vector; is the wave number vector, representing frequency in Fourier space; is the system state vector The first element.

6. The data model system for the limit value of the intelligent reasoning test object according to claim 5, characterized in that: In the limit reasoning part, an initial uniform coarse grid is created based on the discretized parameter space through the following process: Each parameter in , choose an initial coarse grid step size , which is larger than the discretization step size : ； in, is an integer greater than 1; the initial uniform coarse grid It is expressed as: ； The total number of grid points Calculated by the following formula: ； in, is the scaling factor between the initial coarse grid step size and the discretization step size; is a floor function; the coordinates of the grid points are expressed as: ； in .

7. The data model system for the limit value of the intelligent reasoning test object according to claim 6, characterized in that: In the limit reasoning part, the potential failure matrix in the initial uniform coarse grid is found based on the nonlinear dynamic model through the following process: ； in, It is The first The key parameters are time The corresponding potential failure matrix when It is +1 iteration found +1 key parameter in time The corresponding potential failure matrix when It is an iteration parameter that controls the speed of iteration; is the stress field at location and the gradient with time t, indicating the severity of stress change; represents the second-order F-norm.

8. The data model system for the limit value of the intelligent reasoning test object according to claim 7, characterized in that: In the limit reasoning part, based on the potential failure matrix, the Pareto front is found, and the upper limit value in the set corresponding to the Pareto front is used as the upper limit value corresponding to the parameter; the lower limit value in the set corresponding to the Pareto front is used as the lower limit value corresponding to the parameter.

9. A method for implementing the data model system of the limit value of the intelligent reasoning test object according to any one of claims 1 to 8, characterized in that: The method comprises: Step 1: Obtain the key parameters of the test object, determine the variation range of each parameter, construct the parameter space, discretize the parameter space to obtain the discretized parameter space, and then initialize a corresponding starting parameter vector for each parameter in the discretized parameter space; Step 2: Based on the discretized parameter space, the nonlinear equations of the Lorentz-like system are used to perform deterministic modeling on the test object, and random functions are used to perform non-deterministic modeling to characterize random disturbances. The results of the deterministic modeling and the results of the non-deterministic modeling are combined to obtain the nonlinear dynamic model of the test object; Step 3: Create a multi-scale stress field to simulate the external environment. Based on the discretized parameter space, create an initial uniform coarse grid. Based on the nonlinear dynamic model, find the potential failure matrix in the initial uniform coarse grid through multiple iterations. Each potential failure matrix corresponds to a key parameter. For the potential failure matrix, calculate the upper limit value and lower limit value of each key parameter.

Images