AI intelligent valve actuator training method and system based on multi-modal large model

Abstract

The invention relates to the technical field of AI intelligent valve actuator training, in particular to an AI intelligent valve actuator training method and system based on a multi-modal large model, and the system comprises a training processor, a training error analysis unit, a progressive evaluation unit, a simulation evaluation unit, an autonomous training management unit and a training management unit. According to the method, analysis is performed preliminarily from the angle of the training reading rate in the valve actuator training process, on one hand, effectiveness of the constructed valve dynamic response model is ensured, and on the other hand, safety supervision on the valve actuator training process is facilitated, so that normality of the valve actuator is ensured; the method is advantaged in that the valve actuator training failure rate is further reduced, analysis is further carried out from the versatility angle of the valve actuator, the intelligent valve can be accurately controlled by the current valve actuator in different scenes, analysis is carried out from the simulation test angle, and it is guaranteed that the output action result and the actual result are always kept in the error range.

Description

Technical Field

[0001] The present invention relates to the technical field of AI intelligent valve actuator training, and in particular to an AI intelligent valve actuator training method and system based on a multi-modal large model. Background Art

[0002] The valve actuator is a control actuator for opening and closing the valve. Its structure generally includes a power unit, a control unit and an actuator unit. It is generally used in fluid pipelines, especially liquid pipeline systems. In case of emergency, the relevant operators cannot reach the site to open and close the valve in the first time, or the personnel cannot approach the valve site, and the power unit loses power or gas source. The actuator can drive the valve to a pre-set safe state; However, in the prior art, during the training process of the AI ​​intelligent valve actuator, the training status of the AI ​​intelligent valve actuator cannot be accurately analyzed, which leads to an increase in the risk of training errors of the AI ​​intelligent valve actuator. It is not conducive to the precise control of the AI ​​intelligent valve actuator in different scenarios, and the training accuracy of the AI ​​intelligent valve actuator cannot be verified from a simulated perspective, resulting in an increase in the execution error. At the same time, as the AI ​​intelligent valve degrades, the training cycle of the AI ​​intelligent valve actuator cannot be flexibly and autonomously adjusted, resulting in a decrease in the control accuracy of the AI ​​intelligent valve actuator. In view of the above technical defects, a solution is now proposed. Summary of the invention

[0003] The purpose of the present invention is to provide an AI intelligent valve actuator training method and system based on a multimodal large model to solve the above-mentioned technical defects. The present invention initially analyzes from the perspective of the training reading rate during the training process of the AI ​​intelligent valve actuator. On the one hand, it helps to ensure the effectiveness of the constructed valve dynamic response model. On the other hand, it helps to safely supervise the training process of the AI ​​intelligent valve actuator to ensure the normal operation of the AI ​​intelligent valve actuator, thereby reducing the failure rate of the AI ​​intelligent valve actuator training. It further analyzes from the perspective of the versatility of the AI ​​intelligent valve actuator so as to adjust the training range of the current AI intelligent valve actuator so as to improve the current AI intelligent valve actuator. The AI ​​intelligent valve actuator can accurately control the AI ​​intelligent valve in different scenarios. At the same time, it analyzes from the perspective of simulation testing to ensure that the output action result and the actual result are always kept within the error range. The comprehensive performance of the AI ​​intelligent valve actuator is analyzed in an in-depth manner so as to make reasonable autonomous training cycle adjustments according to the performance degradation of the AI ​​intelligent valve to improve the continuous control quality and accuracy of the AI ​​intelligent valve actuator.

[0004] The object of the present invention can be achieved by the following technical solutions: an AI intelligent valve actuator training system based on a multimodal large model, comprising a training processor, a training error analysis unit, a progressive evaluation unit, a simulation evaluation unit, an autonomous training management unit, and a training management unit; The training processor is used to retrieve the historical operation data of the AI ​​intelligent valve actuator, and send the historical operation data to the training error analysis unit for training error tracking evaluation analysis to obtain an abnormal signal or a normal signal. When a normal signal is generated, the progressive evaluation unit is used to perform training discrimination processing on the output result of the adaptive training of the collected AI intelligent valve actuator to obtain a comprehensive signal or a defect signal, wherein the adaptive training includes closed-loop test training and fault tolerance training; When a normal signal is generated, the simulation evaluation unit is used to perform simulation test feedback analysis on the measured error value of the AI ​​intelligent valve actuator to obtain a stable signal or a deviation signal. The autonomous training management unit is used to obtain the comprehensive performance score of the AI ​​intelligent valve, and at the same time, perform autonomous training update cycle regulation and analysis on the comprehensive performance score to obtain the autonomous training planning update cycle.

[0005] Preferably, the training error tracking, evaluation and analysis process is as follows: The training period of the AI ​​intelligent valve actuator is collected, and the training period of the AI ​​intelligent valve actuator is set as the time threshold, and the historical operation data of the AI ​​intelligent valve actuator within the time threshold is obtained. The historical operation data includes pressure, flow, and stroke position. Based on the historical operation data and the LSTM neural network, a valve dynamic response model is constructed; Obtain the training reading rate curve of the AI ​​intelligent valve actuator within the time threshold, obtain the maximum peak value and the minimum trough value from the training reading rate curve, obtain the difference between the maximum peak value and the minimum trough value, and perform discrimination processing on the difference between the maximum peak value and the minimum trough value. If the difference between the maximum peak value and the minimum trough value is less than the preset threshold, generate a feedback instruction; When a feedback instruction is generated, the average value of the maximum peak value and the minimum trough value is obtained, and the average value of the maximum peak value and the minimum trough value is judged and processed to obtain an abnormal signal or a normal signal.

[0006] Preferably, the training and discrimination processing process is as follows: The output results of the adaptive training of the AI ​​intelligent valve actuator within the time threshold are obtained. The adaptive training includes closed-loop test training and fault tolerance training. The output results include qualified and unqualified. The output results of the adaptive training of the AI ​​intelligent valve actuator are obtained as the number corresponding to the qualified ones, and the output results of the adaptive training of the AI ​​intelligent valve actuator are judged and processed to obtain a comprehensive signal or a defect signal.

[0007] Preferably, the simulation test feedback analysis process is as follows: A dynamic physical model is constructed in a simulation environment, and the visual image of the AI ​​smart valve in the current setting environment is obtained. At the same time, the input text instructions are obtained, and guidance instructions based on the visual image and text instructions are generated. The visual image, text instructions and guidance instructions are input into the valve dynamic response model to obtain the action instructions output by the valve dynamic response model, wherein the text instructions are the task objectives of the AI ​​smart valve actuator to perform tasks.

[0008] Preferably, the measured error value between the actual response action instruction of the AI ​​intelligent valve actuator in the dynamic physical model and the action instruction output by the valve dynamic response model is obtained, and the measured error value is judged and processed. If the measured error value is less than or equal to the preset measured error value threshold, a stable signal is generated; if the measured error value is greater than the preset measured error value threshold, a deviation signal is generated.

[0009] Preferably, the autonomous training update cycle regulation and analysis process is as follows: historical operating data and sensor monitoring data of the AI ​​smart valve within the time threshold are obtained, the historical operating data include the number of opening and closing times, torque value, and response time, and the sensor monitoring data include vibration and temperature. The historical operating data and sensor monitoring data are preprocessed, and the preprocessing includes cleaning and data enhancement. A comprehensive performance scoring model is constructed based on the preprocessed historical operating data and sensor monitoring data.

[0010] Preferably, the comprehensive performance score of the AI ​​intelligent valve within the time threshold is obtained based on the comprehensive performance scoring model, and the comprehensive performance score is discriminated. If the comprehensive performance score is less than the preset comprehensive performance score threshold, the change in the comprehensive performance score is obtained, and the change in the comprehensive performance score is discriminated: if the change in the comprehensive performance score is less than the preset threshold, it is determined to be a first-level degradation; if the change in the comprehensive performance score is greater than or equal to the preset threshold, it is determined to be a second-level degradation, and the preset period reduction value corresponding to the first-level degradation or the second-level degradation is obtained, and the autonomous training update period of the AI ​​intelligent valve actuator within the time threshold is obtained, and the value obtained by subtracting the product of the autonomous training update period and the preset period reduction value from the autonomous training update period is set as the autonomous training planning update period.

[0011] The AI ​​intelligent valve actuator training method based on multimodal large model includes the following steps: Step 1: Early stage valve dynamic response model training error evaluation process; Step 2: Evaluate the training scope of the AI ​​intelligent valve actuator from the perspective of versatility through information progression; Step 3: Execution error supervision operation of AI intelligent valve actuator based on simulation environment; Step 4: Construct a comprehensive performance scoring model for AI smart valves and an autonomous training cycle control process based on the analysis of the comprehensive performance scoring model.

[0012] The beneficial effects of the present invention are as follows: (1) The present invention preliminarily analyzes the training reading rate during the training process of the AI ​​intelligent valve actuator. On the one hand, it helps to ensure the effectiveness of the constructed valve dynamic response model. On the other hand, it helps to safely supervise the training process of the AI ​​intelligent valve actuator to ensure the normal operation of the AI ​​intelligent valve actuator, thereby reducing the failure rate of the AI ​​intelligent valve actuator training. The present invention further analyzes the versatility of the AI ​​intelligent valve actuator so as to adjust the training range of the current AI intelligent valve actuator to improve the ability of the current AI intelligent valve actuator to accurately control the AI ​​intelligent valve in different scenarios. (2) The present invention analyzes from the perspective of simulation testing to ensure that the output action results and the actual results are always kept within the error range, and analyzes the comprehensive performance of the AI ​​smart valve in an in-depth manner so as to make reasonable autonomous training cycle adjustments based on the performance degradation of the AI ​​smart valve, so as to improve the continuous control quality and accuracy of the AI ​​smart valve actuator. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention will be further described below in conjunction with the accompanying drawings; Figure 1 It is a flowchart of the system of the present invention; Figure 2 It is a reference analysis diagram of the method of the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] Embodiment 1: See also Figure 1 to Figure 2As shown, the present invention is an AI intelligent valve actuator training system based on a multimodal large model, including a training processor, a training error analysis unit, a progressive evaluation unit, a simulation evaluation unit, an autonomous training management unit and a training management unit, the training processor is connected to the training error analysis unit in a one-way communication, the training error analysis unit is connected to the progressive evaluation unit in a two-way communication, the training error analysis unit is connected to the simulation evaluation unit and the training management unit in a one-way communication, the simulation evaluation unit is connected to the autonomous training management unit and the training management unit in a one-way communication, and the autonomous training management unit is connected to the training management unit in a one-way communication; The training processor is used to retrieve the historical operation data of the AI ​​intelligent valve actuator and send the historical operation data to the training error analysis unit. The training error analysis unit is used to perform training error tracking, evaluation and analysis on the received historical operation data, and to perform discrimination processing on the mean of the sum of the maximum peak value and the minimum trough value obtained to understand whether the construction of the valve dynamic response model is effective, so as to reduce the failure rate of training; The specific training error tracking, evaluation and analysis process is as follows: the training period of the AI ​​intelligent valve actuator is collected, and the training period of the AI ​​intelligent valve actuator is set as the time threshold; Obtain the historical operation data of the AI ​​intelligent valve actuator within the time threshold. The historical operation data includes time series data such as pressure, flow, and stroke position. Based on the historical operation data and the LSTM neural network, a valve dynamic response model is constructed. The training reading rate curve of the AI ​​intelligent valve actuator within the time threshold is obtained, the maximum peak value and the minimum trough value are obtained from the training reading rate curve, the difference between the maximum peak value and the minimum trough value is obtained, and the difference between the maximum peak value and the minimum trough value is discriminated. If the difference between the maximum peak value and the minimum trough value is less than the preset threshold, a feedback instruction is generated. When the feedback instruction is generated, the mean of the maximum peak value and the minimum trough value is obtained, and the mean of the maximum peak value and the minimum trough value is discriminated. If the mean of the maximum peak value and the minimum trough value is less than the preset threshold, an abnormal signal is generated, and it is judged that the construction of the valve dynamic response model is abnormal. If the mean of the maximum peak value and the minimum trough value is greater than or equal to the preset threshold, a normal signal is generated, and it is judged that the construction of the valve dynamic response model is normal. The abnormal signal or the normal signal is sent to the training management unit. After receiving the abnormal signal or the normal signal, the training management unit immediately displays the preset warning text corresponding to the abnormal signal or the normal signal to ensure the effectiveness of the construction of the valve dynamic response model and reduce the failure rate of training.

[0016] Embodiment 2: When a normal signal is generated, the progressive evaluation unit is used to collect the output results of the adaptive training of the AI ​​intelligent valve actuator, and perform training and discrimination processing on the output results, so as to adjust the training range of the current AI intelligent valve actuator, so as to improve the current AI intelligent valve actuator to accurately control the AI ​​intelligent valve in different scenarios, which is helpful to improve the versatility and comprehensiveness of the AI ​​intelligent valve actuator. The specific training and discrimination processing process is as follows: the output results of the adaptive training of the AI ​​intelligent valve actuator within the time threshold are obtained, and the adaptive training includes closed-loop test training, fault tolerance training, etc. The output results include qualified and unqualified, and the number of corresponding qualified output results of the adaptive training of the AI ​​intelligent valve actuator is obtained, and the number of corresponding qualified output results of the adaptive training of the AI ​​intelligent valve actuator is discriminated: If the output result of the adaptive training of the AI ​​smart valve actuator is that the number of qualified correspondences is equal to the preset threshold, a comprehensive signal is generated; If the output result of the adaptive training of the AI ​​intelligent valve actuator is that the number of qualified ones is equal to the preset threshold, a defect signal is generated, and the comprehensive signal or defect signal is sent to the training management unit through the training error analysis unit. After receiving the comprehensive signal or defect signal, the training management unit immediately displays the preset warning text corresponding to the comprehensive signal or defect signal, so as to adjust the training range of the current AI intelligent valve actuator, so as to improve the current AI intelligent valve actuator to accurately control the AI ​​intelligent valve in different scenarios, which is helpful to improve the versatility and comprehensiveness of the AI ​​intelligent valve actuator; Among them, closed-loop test training refers to testing and training the control accuracy of the AI ​​intelligent valve actuator under different load disturbances, and fault tolerance training refers to testing and training the operating status of the AI ​​intelligent valve actuator when the host computer command is lost; When a normal signal is generated, the simulation evaluation unit is used to collect the measured error value of the AI ​​intelligent valve actuator, and perform simulation test feedback analysis on the measured error value of the AI ​​intelligent valve actuator, that is, to analyze from the perspective of simulation execution accuracy to determine whether the training accuracy of the AI ​​intelligent valve actuator is qualified, so as to reduce the error risk by increasing training samples or reducing model complexity. The specific simulation test feedback analysis process is as follows: a dynamic physical model is constructed in a simulation environment, and a visual image of the AI ​​intelligent valve in the current setting environment is obtained, and an input text instruction is obtained at the same time, and a guidance instruction based on the visual image and the text instruction is generated, and the visual image, text instruction and guidance instruction are input into the valve dynamic response model to obtain the action instruction output by the valve dynamic response model, wherein the text instruction is the task goal of the AI ​​intelligent valve actuator to perform the task; It should be noted that the multimodal content consisting of visual images and text instructions is used as input. The visual images give the model the basic function of perceiving the execution environment of the AI ​​smart valve actuator, while the text instructions are the task objectives of the AI ​​smart valve actuator. The guidance instructions of the relevant callable API serve as a communication bridge between natural language and the execution control of the AI ​​smart valve actuator, thereby realizing the control of the action of the AI ​​smart valve actuator by using the language ability of the execution large model; In the process of understanding the task text instructions, multimodal matching is of utmost importance. The AI ​​intelligent valve actuator itself does not have the ability to understand semantics. Therefore, for the object to be operated, the object needs to retrieve the corresponding image through text description, and then further obtain its three-dimensional real space position through methods such as camera calibration, and then hand it over to the AI ​​intelligent valve actuator to perform the operation; The measured error value between the actual response action instruction of the AI ​​intelligent valve actuator in the dynamic physical model and the action instruction output by the valve dynamic response model is obtained, and the measured error value is discriminated and processed. If the measured error value is less than or equal to the preset measured error value threshold, a stable signal is generated. If the measured error value is greater than the preset measured error value threshold, a deviation signal is generated, and the stable signal or the deviation signal is sent to the training management unit. After receiving the stable signal or the deviation signal, the training management unit immediately stabilizes the preset warning text corresponding to the signal or the deviation signal, so as to reduce the error risk by increasing the training samples or reducing the model complexity, thereby helping to improve the control accuracy of the AI ​​intelligent valve actuator, while ensuring that the output action result and the actual result are always kept within the error range.

[0017] Embodiment three: When a stable signal is generated or a comprehensive signal is displayed, the autonomous training management unit is used to obtain the comprehensive performance score of the AI ​​smart valve, and at the same time, the comprehensive performance score is analyzed through autonomous training update cycle regulation, so as to make reasonable training cycle adjustments according to the performance degradation of the AI ​​smart valve, so as to improve the continuous control quality and accuracy of the AI ​​smart valve actuator. The specific autonomous training update cycle regulation and analysis process is as follows: the historical operating data and sensor monitoring data of the AI ​​smart valve within the time threshold are obtained, the historical operating data includes the number of opening and closing times, torque value, response time, etc., the sensor monitoring data includes vibration, temperature, etc., the historical operating data and sensor monitoring data are preprocessed, the preprocessing includes cleaning, data enhancement, etc., and a comprehensive performance scoring model is constructed based on the preprocessed historical operating data and sensor monitoring data; Based on the comprehensive performance scoring model, the comprehensive performance score of the AI ​​intelligent valve within the time threshold is obtained, and the comprehensive performance score is discriminated. If the comprehensive performance score is less than the preset comprehensive performance score threshold, the change in the comprehensive performance score is obtained, and the change in the comprehensive performance score is discriminated: if the change in the comprehensive performance score is less than the preset threshold, it is determined to be a first-level degradation; if the change in the comprehensive performance score is greater than or equal to the preset threshold, it is determined to be a second-level degradation. It should be noted that the degree of performance degradation corresponding to the first-level degradation and the second-level degradation increases in sequence, and the preset cycle reduction value corresponding to the first-level degradation or the second-level degradation is obtained. The autonomous training update cycle of the AI ​​intelligent valve actuator within the time threshold is obtained, and the value obtained by subtracting the product of the autonomous training update cycle and the preset cycle reduction value from the autonomous training update cycle is set as the autonomous training plan update cycle, and the autonomous training plan update cycle is sent to the training management unit. After receiving the autonomous training plan update cycle, the training management unit autonomously trains the AI ​​intelligent valve actuator according to the autonomous training plan update cycle, that is, the training cycle is reasonably adjusted according to the performance degradation of the AI ​​intelligent valve to improve the control quality and accuracy of the AI ​​intelligent valve actuator.

[0018] Embodiment 4: The AI ​​intelligent valve actuator training method based on multimodal large model includes the following steps: Step 1: The early stage of the valve dynamic response model training error evaluation process, that is, to track and evaluate the training error of the historical operation data, and to judge the mean of the maximum peak value and the minimum trough value to determine whether the training of the AI ​​intelligent valve actuator is normal, so as to ensure the effectiveness of the valve dynamic response model; Step 2: Evaluate the training scope of the AI ​​intelligent valve actuator from the perspective of versatility through information progression, that is, perform training discrimination processing on the output results of the adaptive training of the AI ​​intelligent valve actuator to determine whether the versatility and comprehensiveness of the AI ​​intelligent valve actuator training are qualified; Step 3: Based on the execution error supervision operation of the AI ​​intelligent valve actuator in the simulation environment, that is, the simulation test feedback analysis is performed on the measured error value of the AI ​​intelligent valve actuator to determine whether the training accuracy of the AI ​​intelligent valve actuator is qualified; Step 4: Construct a comprehensive performance scoring model for AI intelligent valves and an autonomous training cycle control process based on the analysis of the comprehensive performance scoring model, that is, preprocess and construct a comprehensive performance scoring model based on historical operating data and sensor monitoring data, and perform autonomous training update cycle control analysis based on the comprehensive performance score output by the comprehensive performance scoring model, and autonomously train the valve actuator according to the obtained autonomous training planning update cycle; In summary, the present invention preliminarily analyzes from the perspective of the training reading rate during the training process of the AI ​​intelligent valve actuator. On the one hand, it helps to ensure the effectiveness of the constructed valve dynamic response model. On the other hand, it helps to safely supervise the training process of the AI ​​intelligent valve actuator to ensure the normal operation of the AI ​​intelligent valve actuator, thereby reducing the failure rate of the AI ​​intelligent valve actuator training. It further analyzes from the perspective of the versatility of the AI ​​intelligent valve actuator so as to adjust the training range of the current AI intelligent valve actuator so as to improve the current AI intelligent valve actuator to accurately control the AI ​​intelligent valve in different scenarios. At the same time, it analyzes from the perspective of simulation testing to ensure that the output action result and the actual result are always kept within the error range. The comprehensive performance of the AI ​​intelligent valve is analyzed in an in-depth manner so as to make reasonable autonomous training cycle adjustments according to the performance degradation of the AI ​​intelligent valve so as to improve the continuous control quality and accuracy of the AI ​​intelligent valve actuator.

[0019] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technicians in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0020] The size of the coefficient is to quantify each parameter to obtain a specific value for subsequent comparison. The size of the coefficient depends on the amount of sample data and the preliminary setting of the corresponding operating coefficient for each set of sample data by technical personnel in this field; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0021] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. AI intelligent valve actuator training system based on multi-modal large model, characterized by: It includes a training processor, a training error analysis unit, a progressive evaluation unit, a simulation evaluation unit, an autonomous training management unit, and a training management unit; The training processor is used to retrieve the historical operation data of the AI ​​intelligent valve actuator, and send the historical operation data to the training error analysis unit for training error tracking evaluation analysis to obtain an abnormal signal or a normal signal. When a normal signal is generated, the progressive evaluation unit is used to perform training discrimination processing on the output result of the adaptive training of the collected AI intelligent valve actuator to obtain a comprehensive signal or a defect signal, wherein the adaptive training includes closed-loop test training and fault tolerance training; When a normal signal is generated, the simulation evaluation unit is used to perform simulation test feedback analysis on the measured error value of the AI ​​intelligent valve actuator to obtain a stable signal or a deviation signal. The autonomous training management unit is used to obtain the comprehensive performance score of the AI ​​intelligent valve, and at the same time, perform autonomous training update cycle regulation and analysis on the comprehensive performance score to obtain the autonomous training planning update cycle.

2. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 1 is characterized in that: The training error tracking evaluation and analysis process is as follows: The training period of the AI ​​intelligent valve actuator is collected, and the training period of the AI ​​intelligent valve actuator is set as the time threshold, and the historical operation data of the AI ​​intelligent valve actuator within the time threshold is obtained. The historical operation data includes pressure, flow, and stroke position. Based on the historical operation data and the LSTM neural network, a valve dynamic response model is constructed; Obtain the training reading rate curve of the AI ​​intelligent valve actuator within the time threshold, obtain the maximum peak value and the minimum trough value from the training reading rate curve, obtain the difference between the maximum peak value and the minimum trough value, and perform discrimination processing on the difference between the maximum peak value and the minimum trough value. If the difference between the maximum peak value and the minimum trough value is less than the preset threshold, generate a feedback instruction; When a feedback instruction is generated, the average value of the maximum peak value and the minimum trough value is obtained, and the average value of the maximum peak value and the minimum trough value is judged and processed to obtain an abnormal signal or a normal signal.

3. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 2 is characterized in that: The training and discrimination process is as follows: The output results of the adaptive training of the AI ​​intelligent valve actuator within the time threshold are obtained. The adaptive training includes closed-loop test training and fault tolerance training. The output results include qualified and unqualified. The output results of the adaptive training of the AI ​​intelligent valve actuator are obtained as the number corresponding to the qualified ones, and the output results of the adaptive training of the AI ​​intelligent valve actuator are judged and processed to obtain a comprehensive signal or a defect signal.

4. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 2 is characterized in that: The simulation test feedback analysis process is as follows: A dynamic physical model is constructed in a simulation environment, and a visual image of the AI ​​smart valve in the current setting environment is obtained. At the same time, the input text instructions are obtained, and guidance instructions based on the visual image and text instructions are generated. The visual image, text instructions and guidance instructions are input into the valve dynamic response model to obtain the action instructions output by the valve dynamic response model, wherein the text instructions are the task objectives of the AI ​​smart valve actuator to perform tasks.

5. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 4 is characterized in that: The measured error value between the actual response action instruction of the AI ​​intelligent valve actuator in the dynamic physical model and the action instruction output by the valve dynamic response model is obtained, and the measured error value is judged and processed. If the measured error value is less than or equal to the preset measured error value threshold, a stable signal is generated; if the measured error value is greater than the preset measured error value threshold, a deviation signal is generated.

6. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 2 is characterized in that: The autonomous training update cycle regulation and analysis process is as follows: obtaining the historical operating data and sensor monitoring data of the AI ​​smart valve within the time threshold, the historical operating data including the number of opening and closing times, torque value, and response time, the sensor monitoring data including vibration and temperature, preprocessing the historical operating data and sensor monitoring data, the preprocessing including cleaning and data enhancement, and building a comprehensive performance scoring model based on the preprocessed historical operating data and sensor monitoring data.

7. The AI ​​intelligent valve actuator training system based on multimodal large model according to claim 6 is characterized in that: Based on the comprehensive performance scoring model, the comprehensive performance score of the AI ​​intelligent valve within the time threshold is obtained, and the comprehensive performance score is discriminated. If the comprehensive performance score is less than the preset comprehensive performance score threshold, the change in the comprehensive performance score is obtained, and the change in the comprehensive performance score is discriminated: if the change in the comprehensive performance score is less than the preset threshold, it is judged as primary degradation; if the change in the comprehensive performance score is greater than or equal to the preset threshold, it is judged as secondary degradation, and the preset cycle reduction value corresponding to the primary degradation or secondary degradation is obtained. The autonomous training update cycle of the AI ​​intelligent valve actuator within the time threshold is obtained, and the value obtained by subtracting the product of the autonomous training update cycle and the preset cycle reduction value from the autonomous training update cycle is set as the autonomous training planning update cycle.

8. The AI ​​intelligent valve actuator training method based on a multimodal large model is applied to the AI ​​intelligent valve actuator training system based on a multimodal large model as described in any one of claims 1 to 7, characterized in that: The following steps are involved: Step 1: Early stage valve dynamic response model training error evaluation process; Step 2: Evaluate the training scope of the AI ​​intelligent valve actuator from the perspective of versatility through information progression; Step 3: Execution error supervision operation of AI intelligent valve actuator based on simulation environment; Step 4: Construct a comprehensive performance scoring model for AI smart valves and an autonomous training cycle control process based on the analysis of the comprehensive performance scoring model.

Images