Data processing system for constructing a digital numerical fusion device based on a supercomputer

Abstract

The present invention relates to a data processing system for constructing a digital and numerical fusion device based on a supercomputer, including a cloud computing platform and a supercomputer. The cloud computing platform and the supercomputer are communicatively connected. The supercomputer includes an artificial intelligence model library, a processor, and a memory storing a computer program. The artificial intelligence model library stores a pre-trained autoencoder model. The present invention can quickly and accurately construct a digital and numerical fusion device to achieve full-cycle simulation of equipment.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data processing system for constructing a digital numerical fusion device based on a supercomputer. Background Art

[0002] A digital numerical fusion device is an original simulation model that can construct a virtual physical mapping of a real device or equipment, can display the operating state of the physical device in real time, and reveal the internal operating laws of the physical device. Constructing a digital numerical fusion device is of great significance for the preliminary research and development and finalization of physical devices, virtual process adjustment during operation, and preventive maintenance in the later stage. However, since the digital numerical fusion device itself is strongly coupled with ultra-large-scale computing and requires a huge amount of computing, and because the entire life cycle of many physical devices is long, the management and control of the entire life cycle of physical devices have high requirements for the storage of massive data. If a digital numerical fusion device is directly generated based on the existing data processing system, the operation speed is slow and the construction efficiency is low. When faced with massive data, it is even impossible to complete the construction of the digital numerical fusion device, or some data is lost, and the accuracy of generating the digital numerical fusion device cannot be guaranteed. Therefore, how to provide a computing environment that supports the construction of a digital numerical fusion device and realizes the rapid and accurate construction of a digital numerical fusion device has become an urgent technical problem to be solved. Summary of the Invention

[0003] The purpose of the present invention is to provide a data processing system for constructing a digital numerical fusion device based on a supercomputer, which can rapidly and accurately construct a digital numerical fusion device.

[0004] The present invention provides a data processing system for constructing a digital numerical fusion device based on a supercomputer, including a cloud computing platform and a supercomputer. The cloud computing platform and the supercomputer are communicatively connected. The supercomputer includes an artificial intelligence model library, a processor, and a memory storing a computer program. The artificial intelligence model library stores a pre-trained autoencoder model. When the processor executes the computer program, the following steps are implemented:

[0005] Step S1: Obtain an original simulation model, a sampling sensor data set, and a fixed parameter list {FP1, FP2,... FP M} of the original simulation model based on the cloud computing platform. FPm represents the m-th fixed parameter, and the value range of m is from 1 to M. A variable parameter list {VP1, VP2 … VP N} and actual field data, where the sampled sensor data set includes the correspondence between the sampling sensors and the original simulation model, M represents the total number of fixed parameters of the current original simulation model, N represents the total number of variable parameters in the current original simulation model, VP i represents the i-th variable parameter, VP i =(S i , E i , step i ), S i represents the minimum value of VP i , E i represents the maximum value of VP i , step i represents the adjustment step size of VP i , and the value of i ranges from 1 to N;

[0006] Step S2, based on (S i , E i , step i ), obtain the number of parameter combinations Q of the variable parameters:

[0007] ;

[0008] Step S3, compare Q with the preset combination number threshold D. If Q is greater than D, execute Step S4;

[0009] Step S4, randomly select a candidate variable parameter combination {VP 1j , VP 2j … VP Nj} from the Q parameter combinations, VP ij represents the i-th variable parameter value in the candidate variable parameter combination, and the value of j ranges from 1 to Q;

[0010] Step S5, generate a candidate simulation model from the candidate variable parameter combination, the fixed parameter list, the original simulation model, and the sampled sensor data set;

[0011] Step S6, run the candidate simulation model to obtain candidate simulation field data, input the candidate simulation field data and the actual field data into the autoencoder model, output the credibility, and determine whether the credibility is greater than the preset credibility threshold. If so, determine the candidate simulation model as the digital numerical fusion device and end the process. Otherwise, determine the adjustment direction and the step size adjustment parameter h based on the credibility, and update the candidate variable parameter combination: if the adjustment direction is positive, then let , if the adjustment direction is negative, then let , and return to execute Step S5.

[0012] The present invention has obvious advantages and beneficial effects compared with the prior art. By means of the above technical solution, a data processing system for constructing a digital numerical fusion device based on a supercomputer provided by the present invention can achieve quite remarkable technical progressiveness and practicability, and has wide utilization value in the industry. It has at least the following advantages:

[0013] Based on the supercomputer, the present invention can accurately generate corresponding digital numerical fusion devices for the entire life cycle of the equipment, realize accurate simulation of the entire equipment life cycle. Based on the supercomputer, the entire system can quickly process the massive data generated during the entire process of generating the digital numerical fusion device, and can store the records of the digital numerical fusion device and the corresponding massive data such as the digital numerical fusion device, providing a data basis for the generation of the digital numerical fusion device corresponding to the subsequent life cycle of the equipment.

[0014] The above description is only an overview of the technical solution of the present invention. In order to understand the technical means of the present invention more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following preferred embodiments are specifically given, and in conjunction with the accompanying drawings, the details are described as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of a data processing system for constructing a digital numerical fusion device based on a supercomputer provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific embodiments and their effects of a data processing system for constructing a digital numerical fusion device based on a supercomputer proposed by the present invention.

[0017] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of the steps can be implemented in parallel, concurrently, or simultaneously. The process can be terminated when its operation is completed, but it can also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0018] An embodiment of the present invention provides a data processing system for constructing a digital numerical fusion device based on a supercomputer, as Figure 1As shown, it includes a cloud computing platform and a supercomputer. The cloud computing platform and the supercomputer are communicatively connected. The supercomputer includes an artificial intelligence model library, a processor, and a memory storing a computer program. The artificial intelligence model library stores a pre-trained autoencoder model. When the processor executes the computer program, the following steps are implemented:

[0019] Step S1: Obtain an original simulation model, a sampling sensor data set, a fixed parameter list {FP1, FP2,... FP M} of the original simulation model based on the cloud computing platform, where FPm represents the m-th fixed parameter, m ranges from 1 to M, a variable parameter list {VP1, VP2 … VP N} of the original simulation model, and actual field data. Among them, the sampling sensor data set includes the correspondence between sampling sensors and the original simulation model. M represents the total number of fixed parameters of the current original simulation model, N represents the total number of variable parameters in the current original simulation model, VP i represents the i-th variable parameter, VP i =(S i , E i , step i ), S i represents the minimum value of VP i , E i represents the maximum value of VP i , step i represents the adjustment step size of VP i , and i ranges from 1 to N;

[0020] Among them, the original simulation model can be directly constructed based on existing simulation model construction tools. The simulation model construction tools include CAD, etc., and the present invention does not limit this. The parameters in the fixed parameter list refer to the parameters whose values remain fixed over time, such as the length, width, and height of a certain fixed structure of a device. The variable parameter list refers to the parameters that change over time as the device exists, such as the strength of a certain component of the device. The actual field data is the field data obtained through a physical simulation test bench.

[0021] Step S2: Obtain the number of parameter combinations Q of variable parameters based on (S i , E i , step i ):

[0022] ;

[0023] Step S3: Compare Q with a preset combination number threshold D. If Q is greater than D, then execute Step S4;

[0024] It should be noted that the specific value of the combination quantity threshold D is comprehensively set according to the computing power of the system and the target calculation speed of the actual application.

[0025] Step S4: Randomly select a candidate variable parameter combination {VP 1j , VP 2j … VP Nj} from the Q parameter combinations, where VP ij represents the i-th variable parameter value in the candidate variable parameter combination, and the value of j ranges from 1 to Q;

[0026] Step S5: Generate a candidate simulation model from the candidate variable parameter combination, the fixed parameter list, the original simulation model, and the sampling sensor data set;

[0027] Step S6: Run the candidate simulation model to obtain candidate simulation field data, input the candidate simulation field data and the actual field data into the autoencoder model, output the credibility, and determine whether the credibility is greater than a preset credibility threshold. If so, determine the candidate simulation model as the digital numerical fusion device and end the process. Otherwise, determine the adjustment direction and step size adjustment parameter h based on the credibility, and update the candidate variable parameter combination: if the adjustment direction is positive, let , if the adjustment direction is negative, let , and return to execute Step S5.

[0028] When Q is greater than D, if all the parameter combinations of the variable parameters are run, it will bring a great computational burden to the system and reduce the efficiency of constructing the digital numerical fusion device. Therefore, through Steps S4 - S6, the parameter combination of the variable parameters for the next run can be obtained based on the result of each run, making the parameter combination of the variable parameters closer to the real parameter combination of the variable parameters, reducing the computational amount of data processing, and improving the efficiency and accuracy of constructing the digital numerical fusion device.

[0029] The embodiment of the present invention can accurately generate a corresponding digital numerical fusion device for the entire life cycle of the device based on a supercomputer, realize accurate simulation of the entire device life cycle, and based on the supercomputer, enable the entire system to quickly process the massive data in the entire process of generating the digital numerical fusion device, and can store the records of the digital numerical fusion device and the corresponding massive data such as the digital numerical fusion device, providing a data basis for the generation of the digital numerical fusion device corresponding to the subsequent life cycle of the device.

[0030] As an embodiment, when the processor executes the computer program, the following steps are further implemented:

[0031] Step S10: Establish an autoencoder model according to a preset multi - group of measurement point data;

[0032] Step S20: Train the autoencoder model according to the credibility index of the signal reconstruction result.

[0033] Specifically, a physical experimental device model can be established according to experimental requirements, and a preset multi-group of measuring point data can be obtained based on the physical experimental device model. The measuring point data is the detection point data collected by the detection points set on the physical experimental device model. Train the autoencoder model according to the credibility index of the signal reconstruction result, so that the autoencoder model can judge whether the candidate simulation field data and the actual field data match and judge the credibility of the candidate simulation field data.

[0034] When Q is less than or equal to D, it indicates that the computing power of the current system is sufficient to support the operation of directly selecting the optimal data from Q parameter combinations. Therefore, as an embodiment, in step S3, when Q is less than or equal to D, the following steps are performed:

[0035] Step S7: Generate corresponding candidate simulation models based on each parameter combination in the Q parameter combinations, the fixed parameter list, the original simulation model, and the sampling sensor data set;

[0036] Step S8: Run each candidate simulation model to obtain corresponding candidate simulation field data, determine the optimal candidate simulation field data based on the actual field data, and determine the candidate simulation model that outputs the optimal candidate simulation field data as the digital numerical fusion device.

[0037] It can be understood that through steps S1 - S8, the corresponding calculation method can be selected according to the number of parameter combinations and the computing power of the system to obtain the optimal variable parameter combination, improving the construction efficiency of the digital numerical fusion device.

[0038] As an embodiment, in step S8: determining the optimal candidate simulation field data based on the actual field data includes:

[0039] Step S81: Convert the actual field data and each candidate simulation field data into corresponding actual field data vectors and each candidate simulation field data vector respectively;

[0040] Step S82: Obtain the Euclidean distance between each candidate simulation field data vector and the actual field data vector, and determine the candidate simulation field data corresponding to the candidate simulation field data vector with the smallest Euclidean distance as the optimal candidate simulation field data.

[0041] As an embodiment, the supercomputer further includes a storage database. When determining the digital numerical fusion device, it further includes:

[0042] Step S100: Assign a device ID to the digital numerical fusion device, obtain the corresponding user ID, original simulation model ID, fixed parameter list, variable parameter combination, and generation time, generate a digital numerical fusion device record, and store the digital numerical fusion device record and the digital numerical fusion device in the storage database.

[0043] Through step S100, it is possible to store a large amount of important data involved in the construction process of the digital numerical fusion device, providing a data basis for constructing the corresponding digital numerical fusion device for the subsequent life cycle of the device. When constructing the digital numerical fusion device corresponding to the subsequent life cycle, as an embodiment, when the processor executes the computer program, the following steps are also implemented:

[0044] Step S101: Obtain the user ID, original simulation model ID, and current actual field data based on the cloud computing platform;

[0045] Step S102: Retrieve the storage database based on the user ID and original simulation model ID obtained from the cloud computing platform, obtain the digital numerical fusion device closest to the current moment, and obtain the current simulation field data;

[0046] Step S103: Determine whether the error between the current simulation field data and the current actual field data is within the preset error range. If it is, determine the digital numerical fusion device closest to the current moment as the current digital numerical fusion device; otherwise, obtain the fixed parameter list and variable parameter combination closest to the current moment;

[0047] Among them, if the error between the current simulation field data and the current actual field data is within the preset error range, it means that the current digital numerical fusion device still meets the simulation requirements of the current device state. Then, directly determine the digital numerical fusion device closest to the current moment as the digital numerical fusion device corresponding to the current device state, saving computing resources.

[0048] Step S104: Obtain the current variable parameter list and current fixed parameter list based on the cloud computing platform, and generate a target fixed parameter list and a target variable parameter list by combining the fixed parameter list and variable parameter combination closest to the current moment;

[0049] It should be noted that due to factors such as damage to some components of the device or the physical characteristics of some components, the fixed parameter list and variable parameters corresponding to different life periods may change. Therefore, the current variable parameter list and current fixed parameter list are dynamically adjusted according to the fixed parameter list and variable parameter combination closest to the current moment to generate a target fixed parameter list and a target variable parameter list, improving the data processing efficiency and accuracy of constructing the digital numerical fusion device.

[0050] Step S105: Use the target variable parameter list as the fixed parameter list of the original simulation model, and use the target fixed parameter list as the variable parameter list of the original simulation model. Then execute Step S2 to Step S6 or Step S2 to Step S7 to determine the corresponding target digital numerical fusion device.

[0051] As an embodiment, when determining the target digital numerical fusion device, execute Step S100, that is, the target digital numerical fusion device assigns a device id, obtains the corresponding user id, original simulation model id, fixed parameter list, variable parameter combination, and generation time, and generates a digital numerical fusion device record. Store the digital numerical fusion device record and the digital numerical fusion device in the storage database. In this way, storing each generated digital numerical fusion device in the storage database can achieve full coverage of the device cycle, enabling each life stage of the device to correspond to an accurate and reliable digital numerical fusion device, thus realizing device simulation. It can be understood that due to factors such as cost, the simulation range of the physical test bench is limited. However, based on the digital numerical fusion device, any number of measurement points can be arranged at any position for simulation measurement, with low cost and the ability to quickly and accurately obtain simulation results.

[0052] As an embodiment, the cloud computing platform and the supercomputer are communicatively connected based on a preset unified resource call interface, and the unified resource call interface includes a computing interface, a storage interface, a visualization interface, a user management interface, and a software service interface.

[0053] The above is only a preferred embodiment of the present invention and does not impose any form of limitation on the present invention. Although the present invention has been disclosed above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art, without departing from the technical solution of the present invention, can make some changes or modifications to the above-disclosed technical content to obtain equivalent embodiments with equivalent changes. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A data processing system for constructing a digital numerical fusion device based on a supercomputer, characterized in that it includes a cloud computing platform and a supercomputer, the cloud computing platform and the supercomputer are communicatively connected, the supercomputer includes an artificial intelligence model library, a processor, and a memory storing a computer program, and the artificial intelligence model library stores a pre-trained autoencoder model. When the processor executes the computer program, the following steps are implemented: Step S1. Obtain the original simulation model, the sampling sensor data set, the fixed parameter list {FP1, FP2, … FP M}, where FP m represents the m-th fixed parameter, m ranges from 1 to M, the variable parameter list {VP1, VP 2… VP N} of the original simulation model, and the actual field data. Among them, the sampling sensor data set includes the correspondence between the sampling sensors and the original simulation model. M represents the total number of fixed parameters of the current original simulation model, N represents the total number of variable parameters in the current original simulation model, and VP i represents the i-th variable parameter, VP i =(S i , E i , step i ), where S i represents the minimum value of VP i , E i represents the maximum value of VP i , and step i represents the adjustment step of VP i . i ranges from 1 to N; Step S2. Based on (S i , E i , step i ), obtain the number of parameter combinations Q of the variable parameter: ; Step S3: Compare Q with a preset combined quantity threshold D. If Q is greater than D, execute Step S4; Step S4: Randomly select a candidate variable parameter combination {VP 1j , VP 2j… VP Nj} from the Q parameter combinations, where VP ij represents the i-th variable parameter value in the candidate variable parameter combination, and the value of j ranges from 1 to Q; Step S5: Generate a candidate simulation model from the candidate variable parameter combination, the fixed parameter list, the original simulation model, and the sampling sensor data set; Step S6: Run the candidate simulation model to obtain candidate simulation field data. Input the candidate simulation field data and the actual field data into the autoencoder model, and output the credibility. Determine whether the credibility is greater than the preset credibility threshold. If so, determine the candidate simulation model as the digital numerical fusion device and end the process. Otherwise, determine the adjustment direction and the step size adjustment parameter h based on the credibility, and update the candidate variable parameter combination: If the adjustment direction is positive, let VP ij =VP ij + h×step i , if the adjustment direction is negative, let VP ij =VP ij - h×step i , and return to execute Step S5.

2. The system according to claim 1, characterized in that when the processor executes the computer program, the following steps are also implemented: Step S10: Establish an autoencoder model according to a preset multi-group of measurement point data, where the measurement point data is the detection point data collected by the detection points set on the physical experiment device model; Step S20: Train the autoencoder model according to the credibility index of the signal reconstruction result.

3. The system according to claim 1, characterized in that in Step S3, when Q is less than or equal to D, the following steps are executed: Step S7: Generate a corresponding candidate simulation model from each parameter combination in the Q parameter combinations, the fixed parameter list, the original simulation model, and the sampling sensor data set; Step S8: Run each candidate simulation model to obtain the corresponding candidate simulation field data, determine the optimal candidate simulation field data based on the actual field data, and determine the candidate simulation model that outputs the optimal candidate simulation field data as the digital numerical fusion device.

4. The system according to claim 3, characterized in that in Step S8: Determining the optimal candidate simulation field data based on the actual field data includes: Step S81: Convert the actual field data and each candidate simulation field data into corresponding actual field data vectors and each candidate simulation field data vector respectively; Step S82: Obtain the Euclidean distance between each candidate simulation field data vector and the actual field data vector, and determine the candidate simulation field data corresponding to the candidate simulation field data vector with the smallest Euclidean distance as the optimal candidate simulation field data.

5. The system according to claim 1 or 3, characterized in that the supercomputer further includes a storage database. When determining the digital numerical fusion device, it further includes: Step S100: Assign a device id to the digital numerical fusion device, obtain the corresponding user id, original simulation model id, fixed parameter list, variable parameter combination, and generation time, and generate a digital numerical fusion device record. Store the digital numerical fusion device record and the digital numerical fusion device in the storage database.

6. The system according to claim 5, characterized in that when the processor executes the computer program, the following steps are also implemented: Step S101: Obtain the user id, original simulation model id, and current actual field data based on the cloud computing platform; Step S102: Retrieve the storage database based on the user ID and the original simulation model ID obtained from the cloud computing platform, obtain the digital numerical fusion device closest to the current moment, and obtain the current simulation field data; Step S103: Determine whether the error between the current simulation field data and the current actual field data is within a preset error range. If it is, determine the digital numerical fusion device closest to the current moment as the current digital numerical fusion device. Otherwise, obtain the fixed parameter list and variable parameter combination closest to the current moment; Step S104: Obtain the current variable parameter list and the current fixed parameter list based on the cloud computing platform, and generate a target fixed parameter list and a target variable parameter list in combination with the fixed parameter list and variable parameter combination closest to the current moment; Step S105: Use the target variable parameter list as the fixed parameter list of the original simulation model, use the target fixed parameter list as the variable parameter list of the original simulation model, and execute Step S2 to Step S6 or Step S2 to Step S7 to determine the corresponding target digital numerical fusion device.

7. The system according to claim 6, wherein: When determining the target digital numerical fusion device, execute Step S100.

8. The system according to claim 1, wherein: The cloud computing platform and the supercomputer are communicatively connected based on a preset unified resource call interface, and the unified resource call interface includes a computing interface, a storage interface, a visualization interface, a user management interface, and a software service interface.

Images