AI-driven batch complex system model low-redundancy simulation calculation method
By employing an AI-driven low-redundancy simulation computation method for batch complex system models, and using PyTorch and LSTM neural networks to train models and combining them with the Moderna language to construct simulation models, the problem of efficient simulation computation for large-scale system simulation model clusters is solved. This achieves efficient simulation computation and data analysis, thereby improving simulation efficiency and design efficiency.
Patent Information
- Application Number
- CN202411408037.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2025-01-10
AI Technical Summary
Existing technologies struggle to efficiently perform automated simulation and evaluation of large-scale system simulation model clusters within limited time and resources, resulting in low computational efficiency and an inability to quickly obtain system simulation model solutions that best meet the requirements.
We employ an AI-driven low-redundancy simulation computation method for batch complex system models. We train the model using the PyTorch deep learning framework and LSTM neural network, construct the simulation model using the Moderna language, perform low-redundancy computation through neural networks and conditional expressions, and set termination rules to parallelize the computation task, thus forming an efficient simulation solution engine.
It enables efficient simulation calculation and data analysis in scenarios with large number of model clusters, improves simulation calculation efficiency, enables rapid multi-dimensional comprehensive performance optimization, and enhances overall demonstration and design efficiency.
Smart Images
Figure CN119294252A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of engineering system modeling and simulation, and specifically to an AI-driven low-redundancy simulation calculation method for batch complex system models. Background Art
[0002] A system simulation model cluster is based on a certain system simulation model. By arranging and combining model components and their connection relationships, a variety of model connection schemes are derived. Based on this, all possible design value states of key parameters are combined to form a large-scale parameter scheme set, which is integrated into the simulation model. All possible technical states and capabilities of the design space are seamlessly combined to generate a large-scale system simulation model set that can cover the design space.
[0003] A large-scale system simulation model cluster has been formed for the equipment design space, which is usually a collection of models in the tens of thousands. It is necessary to automatically simulate and evaluate the model cluster under limited time and resources. Therefore, it is necessary to improve computing power, optimize the calculation process, and improve automation efficiency to achieve rapid simulation calculation of the model cluster, and perform multi-dimensional comprehensive performance evaluation and optimization according to different evaluation rules to obtain the system simulation model solution that meets the needs to the highest degree. Therefore, it is necessary to study simulation analysis for model clusters, solve computing and data-intensive problems, support the demonstration and design of new equipment, and lay a technical foundation for the formation of advanced overall demonstration and design methods.
[0004] Therefore, technicians in this field have provided an AI-driven low-redundancy simulation calculation method for batch complex system models, which performs simulation calculations on large batches of complex system models, reduces unnecessary simulations during the calculation process, and optimizes the solution process to achieve a simulation model solution for the generated system. Summary of the invention
[0005] 1. Technical issues to be resolved
[0006] In view of the shortcomings of the prior art, the present invention provides an AI-driven low-redundancy simulation calculation method for batch complex system models, forms an efficient simulation solution engine for model clusters, improves the efficiency of system simulation calculation and data analysis in large-scale model cluster scenarios, quickly carries out multi-dimensional comprehensive performance optimization, and improves the efficiency of overall demonstration and design.
[0007] (II) Technical solution
[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0009] An AI-driven low-redundancy simulation calculation method for batch complex system models includes the following steps:
[0010] Step 1. Import large sample data for neural network model training from database or csv file;
[0011] Step 2. Based on the Pytorch deep learning framework, use the LSTM long short-term memory neural network and other related algorithms to train the large sample data in step 1 to form a neural network model and fix the neural network weight parameters;
[0012] Step 3. Encapsulate the neural network model trained in step 2 into a pt file for low-redundancy simulation calculation based on neural network;
[0013] Step 4. Import the large-scale system simulation model built based on the Model lica language;
[0014] Step 5. Based on the model characterization calculation principle, feature recognition is performed on the simulation model in step 4, and similar system simulation models in the model cluster are merged to form a master simulation running instance;
[0015] Step 6. Perform low-redundancy calculation on a single system simulation model;
[0016] Step 7. Perform low-redundancy simulation calculations on similar model clusters;
[0017] Step 8. Generate a result report.
[0018] Furthermore, in step 1, the large sample data is obtained by batch simulation calculation of the large-scale system simulation model constructed based on the Model lica modeling language, injecting the design parameter values of multiple model attributes as simulation input, and marking the unreasonable results in the data.
[0019] Furthermore, in step 5, simulation calculations can be directly performed on the same type of system simulation model clusters to reduce unnecessary compilation processes.
[0020] Furthermore, in step 6, performing low-redundancy calculation on a single system simulation model includes two calculation methods: low-redundancy calculation based on a neural network model, and low-redundancy calculation based on a conditional expression;
[0021] In the low-redundancy calculation based on the neural network model, the neural network model file in step 3 is imported, the simulation process data of the system simulation model in the model cluster is used as input, the system design rationality information is output, and the model calculation is terminated when the termination calculation rule is met;
[0022] In low-redundancy calculations based on conditional expressions, multiple relational expressions are set to express the logical relationships among model parameters, simulation variables, design indicators, expressions, etc., perform preliminary mathematical operations on the relational expressions, and calculate whether multiple expressions meet the preset conditions. The values of the relational expressions are monitored to automatically determine whether the termination calculation rules are met.
[0023] Furthermore, in step 7, the termination calculation rules for the same type of model clusters are set, and the parallelization of the algebraic equation solver based on task scheduling, multi-mode fusion and computing task monitoring provided by the built-in model cluster parallel computing technology are combined to accelerate the solution process. When the preset conditions are met, the simulation of all simulation models in this type of model cluster is terminated immediately.
[0024] Further, an AI-driven batch complex system model includes a model cluster low-redundancy simulation module and a model cluster parallel computing module;
[0025] The model cluster low-redundancy simulation module provides intelligent pre-calculation based on neural networks, matrix processing of simulation running instances, and low-redundancy simulation based on neural networks and conditional expressions to improve simulation efficiency;
[0026] The model cluster parallel computing module performs parallel computing at the system simulation model solution level and large-scale model cluster computing task level, accelerating the simulation operation process of large-scale model clusters.
[0027] Furthermore, the model cluster low-redundancy simulation module includes neural network model training, automatic model arrangement based on model features, generation of master simulation running instances, low-redundancy simulation based on neural networks and low-redundancy simulation based on conditional expressions.
[0028] Furthermore, the model cluster parallel computing module includes parallelization of algebraic equation solvers based on task scheduling, parallel computing based on multi-mode fusion, and computing task monitoring and result analysis.
[0029] (III) Beneficial effects
[0030] The present invention provides an AI-driven low-redundancy simulation calculation method for batch complex system models. It has the following beneficial effects:
[0031] 1. The present invention provides an AI-driven low-redundancy simulation calculation method for batch complex system models, which adopts the standard Modelica language without modifying the syntax of Modelica. Modelica language is an object-oriented, equation-based, non-causal, multi-domain unified modeling language, which has natural advantages in the construction of large-scale complex heterogeneous models involving multiple disciplines and multi-professional coupling such as mechanics, electronics, control, hydraulics, pneumatics, and heat.
[0032] 2. The present invention provides an AI-driven low-redundancy simulation calculation method for batch complex system models. Through the model cluster low-redundancy simulation technology, from the perspective of a single system simulation model and a similar model cluster, it provides an intelligent pre-calculation and simulation calculation method based on a neural network to reduce unnecessary redundant calculations.
[0033] 3. The present invention provides an AI-driven low-redundancy simulation calculation method for batch complex system models, which performs parallel calculations from the system simulation model solution level and the large-scale model cluster calculation task level, accelerates the solution of new model cluster simulation tasks, and forms an efficient simulation solution engine for model clusters, thereby accelerating the efficiency of system simulation calculations and data analysis in model cluster scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 A flow chart of a model low-redundancy simulation calculation method according to an embodiment of the present invention;
[0035] Figure 2 This is an interface diagram of importing large sample data according to an embodiment of the present invention;
[0036] Figure 3 This is a diagram of the training neural network model interface of an embodiment of the present invention;
[0037] Figure 4 A diagram of importing a large number of system simulation model files according to an embodiment of the present invention;
[0038] Figure 5 This is a schematic diagram of the composition of the AI-driven batch complex system model according to an embodiment of the present invention. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the specific embodiments of the present invention to clearly and completely describe the technical solutions in the specific embodiments of the present invention. Obviously, the specific embodiments described are only part of the specific embodiments of the present invention, not all of the specific embodiments. Based on the specific embodiments of the present invention, all other specific embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0041] like Figure 1 As shown, a specific embodiment of the present invention provides an AI-driven batch complex system model low-redundancy simulation calculation method, comprising the following steps:
[0042] Step 1. Import large sample data for neural network model training from a database or csv file. The large sample data consists of a large batch system simulation model built based on the Mode lica modeling language. The design parameter values of multiple model attributes are injected as simulation inputs. The result data obtained by batch simulation calculation is then marked with unreasonable results in the data.
[0043] Step 2. Based on the Pytorch deep learning framework, use the LSTM long short-term memory neural network and other related algorithms to train the large sample data in step 1 to form a neural network model and fix the neural network weight parameters;
[0044] Step 3. Encapsulate the neural network model trained in step 2 into a pt file for low-redundancy simulation calculation based on neural network;
[0045] Step 4. Import the large-scale system simulation model built based on the Model lica language;
[0046] Step 5. Based on the model characterization calculation principle, feature recognition is performed on the simulation model in step 4, and similar system simulation models in the model cluster are merged to form a parent simulation running instance. For the similar system simulation model cluster, simulation calculation can be performed directly to reduce unnecessary compilation process;
[0047] Step 6. Perform low-redundancy calculation on a single system simulation model, including two calculation methods: low-redundancy calculation based on a neural network model, and low-redundancy calculation based on a conditional expression;
[0048] In the low-redundancy calculation based on the neural network model, the neural network model file in step 3 is imported, the simulation process data of the system simulation model in the model cluster is used as input, the system design rationality information is output, and the model calculation is terminated when the termination calculation rule is met;
[0049] In low-redundancy calculation based on conditional expressions, multiple relational expressions are set to express the logical relationship between model parameters, simulation variables, design indicators, expressions, etc., perform preliminary mathematical operations on the relational expressions, and calculate whether multiple expressions meet the preset conditions, monitor the values of the relational expressions, and automatically determine whether the termination calculation rules are met;
[0050] Step 7. Perform low-redundancy simulation calculations on the same model clusters, set termination calculation rules for the same model clusters, and combine the parallelization of algebraic equation solvers based on task scheduling, multi-mode fusion, and computing task monitoring provided by the built-in model cluster parallel computing technology to accelerate the solution process. When the preset conditions are met, immediately terminate the simulation of all simulation models in the same model cluster;
[0051] Step 8. Generate a result report.
[0052] The following is a practical example:
[0053] Step 1. Import large sample data, as shown in the attached Figure 2 As shown;
[0054] Step 2. Train the neural network model, as shown in the attached Figure 3 As shown;
[0055] Step 3. Import the large batch system simulation model, as shown in the attached Figure 4 As shown;
[0056] Step 4. Extract features and merge the simulation models of the same model cluster to generate a master simulation running instance;
[0057] Step 5. Select the system simulation model, set the simulation calculation method and termination conditions, and run the low-redundancy simulation calculation process;
[0058] Step 6. Select the model cluster and simulation calculation method, set the termination condition, perform low-redundancy simulation calculation on it, and obtain the rationality information of system design;
[0059] Step 7. Generate a result report.
[0060] As attached Figure 5 As shown, the AI-driven batch complex system model includes a model cluster low-redundancy simulation module and a model cluster parallel computing module;
[0061] The model cluster low-redundancy simulation module provides intelligent pre-calculation based on neural networks, matrix processing of simulation running instances, and low-redundancy simulation based on neural networks and conditional expressions to improve simulation efficiency;
[0062] The model cluster parallel computing module performs parallel computing at the system simulation model solving level and large-scale model cluster computing task level, accelerating the simulation operation process of large-scale model clusters;
[0063] The model cluster low-redundancy simulation module includes neural network model training, automatic model arrangement based on model features, generation of master simulation running instances, low-redundancy simulation based on neural networks, and low-redundancy simulation based on conditional expressions. The model cluster parallel computing module includes parallelization of algebraic equation solvers based on task scheduling, parallel computing based on multi-mode fusion, and computing task monitoring and result analysis.
[0064] Although specific embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the specific embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. An AI-driven low-redundancy simulation calculation method for batch complex system models, characterized in that: The following steps are involved: Step 1. Import large sample data for neural network model training from database or csv file; Step 2. Based on the Pytorch deep learning framework, use the LSTM long short-term memory neural network and other related algorithms to train the large sample data in step 1 to form a neural network model and fix the neural network weight parameters; Step 3. Encapsulate the neural network model trained in step 2 into a pt file for low-redundancy simulation calculation based on neural network; Step 4. Import the large-scale system simulation model built based on Modelica language; Step 5. Based on the model characterization calculation principle, feature recognition is performed on the simulation model in step 4, and similar system simulation models in the model cluster are merged to form a master simulation running instance; Step 6. Perform low-redundancy calculation on a single system simulation model; Step 7. Perform low-redundancy simulation calculations on similar model clusters; Step 8. Generate a result report.
2. According to claim 1, an AI-driven low-redundancy simulation calculation method for batch complex system models is characterized in that: In step 1, a large sample of data is obtained by batch simulation calculation of a large-scale system simulation model built based on the Modelica modeling language, and design parameter values of multiple model attributes are injected as simulation inputs, and unreasonable results in the data are marked.
3. According to claim 1, an AI-driven batch complex system model low-redundancy simulation calculation method is characterized in that: In step 5, simulation calculations can be directly performed on the same type of system simulation model clusters to reduce unnecessary compilation processes.
4. According to claim 1, an AI-driven batch complex system model low-redundancy simulation calculation method is characterized in that: In step 6, performing low-redundancy calculation on a single system simulation model includes two calculation methods: low-redundancy calculation based on a neural network model, and low-redundancy calculation based on a conditional expression; In the low-redundancy calculation based on the neural network model, the neural network model file in step 3 is imported, the simulation process data of the system simulation model in the model cluster is used as input, the system design rationality information is output, and the model calculation is terminated when the termination calculation rule is met; In low-redundancy calculations based on conditional expressions, multiple relational expressions are set to express the logical relationships among model parameters, simulation variables, design indicators, expressions, etc., perform preliminary mathematical operations on the relational expressions, and calculate whether multiple expressions meet the preset conditions. The values of the relational expressions are monitored to automatically determine whether the termination calculation rules are met.
5. According to claim 1, an AI-driven batch complex system model low-redundancy simulation calculation method is characterized in that: In step 7, the termination calculation rules for the same type of model clusters are set, and the parallelization of the algebraic equation solver based on task scheduling, multi-mode fusion and computing task monitoring provided by the built-in model cluster parallel computing technology are combined to accelerate the solution process. When the preset conditions are met, the simulation of all simulation models in this type of model cluster is terminated immediately.
6. An AI-driven batch complex system model according to any one of claims 1-6, characterized in that: It includes a model cluster low-redundancy simulation module and a model cluster parallel computing module; The model cluster low-redundancy simulation module provides intelligent pre-calculation based on neural networks, matrix processing of simulation running instances, and low-redundancy simulation based on neural networks and conditional expressions to improve simulation efficiency; The model cluster parallel computing module performs parallel computing at the system simulation model solution level and large-scale model cluster computing task level, accelerating the simulation operation process of large-scale model clusters.
7. The AI-driven batch complex system model according to claim 6, characterized in that: The model cluster low-redundancy simulation module includes neural network model training, model automatic arrangement based on model features, master simulation running instance generation, neural network-based low-redundancy simulation and conditional expression-based low-redundancy simulation.
8. The AI-driven batch complex system model according to claim 6, characterized in that: The model cluster parallel computing module includes parallelization of algebraic equation solvers based on task scheduling, parallel computing based on multi-mode fusion, and computing task monitoring and result analysis.