Servo system adaptive parameter adjustment optimization method and system based on machine learning

Abstract

The present invention provides a method and system for adaptive parameter adjustment and optimization of a servo system based on machine learning, which relates to the technical field of machine learning. The method first uses a target servo instruction set and its corresponding priori servo instruction set, and generates an adaptive optimization instruction set that supplements the missing instruction element data through a first model learning module of a machine learning network model, thereby improving the integrity and accuracy of the instruction set. Furthermore, through a second model learning module of the machine learning network model, a servo system parameter configuration result that is highly matched with the target servo working application scenario is generated according to the adaptive optimization instruction set. Finally, based on the parameter configuration result, the servo system is adaptively adjusted and optimized, which significantly improves the performance and stability of the servo system in different working scenarios, realizes the intelligentization and automation of the servo system parameter adjustment, and greatly improves the operating efficiency and accuracy of the servo system.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a method and system for adaptive parameter adjustment and optimization of a servo system based on machine learning. Background Art

[0002] As a core component in the field of automation control, servo systems are widely used in industrial manufacturing, robot control, aerospace and other fields. Traditional servo system parameter adjustment usually relies on engineers' experience and manual debugging. This method is not only time-consuming and labor-intensive, but also difficult to ensure the optimality of parameter adjustment. Especially in complex and changeable working scenarios, the performance of servo systems is often limited.

[0003] With the rapid development of machine learning technology, its advantages in data processing, pattern recognition and decision optimization are gradually emerging. However, in the field of servo system parameter adjustment, how to effectively use machine learning technology to achieve adaptive parameter optimization is still an urgent problem to be solved. Traditional machine learning methods are often directly applied to the control strategy optimization of servo systems, while ignoring the processing and optimization of the servo instruction set itself. As a result, in practical applications, the optimization effect of machine learning models is limited due to the incompleteness of the instruction set or the lack of instruction element data.

[0004] In addition, the working scenarios of servo systems are usually diverse and complex, and different working scenarios have different requirements for the parameter configuration of servo systems. Therefore, how to generate a servo system parameter configuration result that is highly matched with the target servo working application scenario is also a major challenge facing the current servo system parameter adjustment field. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a servo system adaptive parameter adjustment optimization method based on machine learning, the method comprising:

[0006] Acquire a target servo instruction set of a servo system, and a priori servo instruction set corresponding to the target servo instruction set, wherein the target servo instruction set and the priori servo instruction set are servo instruction sets for target servo working application scenarios, and the priori servo instruction set and the target servo instruction set have instruction association features;

[0007] The target servo instruction set and the priori servo instruction set are loaded into a machine learning network model, and an adaptive optimization instruction set corresponding to the target servo instruction set is generated by using a first model learning module of the machine learning network model, wherein the first model learning module of the machine learning network model is used to supplement the missing instruction element data in the target servo instruction set by using optimization;

[0008] The adaptive optimization instruction set is loaded into the machine learning network model, and a servo system parameter configuration result of the adaptive optimization instruction set is generated by using a second model learning module of the machine learning network model, wherein the servo system parameter information included in the servo system parameter configuration result matches the target servo working application scenario;

[0009] The servo system is adaptively parameter-adjusted and optimized based on the servo system parameter configuration result of the adaptive optimization instruction set.

[0010] On the other hand, an embodiment of the present invention also provides a servo system adaptive parameter adjustment and optimization system based on machine learning, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present application realizes efficient and accurate optimization of the servo system parameters by introducing a machine learning network model. Specifically, the method first utilizes the target servo instruction set and its corresponding prior servo instruction set, and generates an adaptive optimization instruction set that supplements the missing instruction element data through the first model learning module of the machine learning network model, thereby improving the completeness and accuracy of the instruction set. Furthermore, through the second model learning module of the machine learning network model, a servo system parameter configuration result that is highly matched with the target servo working application scenario is generated according to the adaptive optimization instruction set. Finally, based on the parameter configuration result, the servo system is adaptively adjusted and optimized, which significantly improves the performance and stability of the servo system in different working scenarios, realizes the intelligent and automated adjustment of the servo system parameters, and greatly improves the operating efficiency and accuracy of the servo system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the servo system adaptive parameter adjustment optimization method based on machine learning provided in an embodiment of the present invention.

[0013] Figure 2 It is a schematic diagram of the hardware architecture of a servo system adaptive parameter adjustment and optimization system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 It is a flow chart of a servo system adaptive parameter adjustment and optimization method based on machine learning provided by an embodiment of the present invention. The servo system adaptive parameter adjustment and optimization method based on machine learning is introduced in detail below.

[0015] Step S110, obtaining a target servo instruction set of a servo system and a priori servo instruction set corresponding to the target servo instruction set, wherein the target servo instruction set and the priori servo instruction set are servo instruction sets for target servo working application scenarios, and the priori servo instruction set has instruction association features with the target servo instruction set.

[0016] In this embodiment, for a servo system on an industrial automation production line, the target servo work application scenario is to perform high-precision processing operations on a precision component. In this scenario, the target servo instruction set includes a series of control instructions for the servo motor, such as precise setting instructions for the motor speed, control instructions for the motor rotation direction, torque output limit instructions, etc. These instructions are all to ensure that each processing action can be accurately executed when processing this precision component.

[0017] The prior servo instruction set is accumulated from previous similar processing operations or the processing of similar parts on the same production line. For example, the servo instruction set used when processing a part with a similar shape but slightly lower precision requirements. Due to the similarity of parts and the partial overlap of processing processes, this prior servo instruction set has instruction association characteristics with the current target servo instruction set. This association feature is reflected in many aspects. For example, in processing path planning, although the accuracy requirements are different, the basic processing sequence and trajectory are similar, so there may be certain similarities in the instructions for controlling the direction of motor rotation and speed change; in terms of torque control, due to the partial similarity of materials, the control instructions for the force applied to the tool during processing may also be associated. By obtaining these target servo instruction sets and prior servo instruction sets, basic data is provided for subsequent optimization processing, so that the existing experience data (prior servo instruction set) can be used to optimize the current target servo instruction set to better meet the needs of the target servo work application scenario.

[0018] Step S120, loading the target servo instruction set and the prior servo instruction set into the machine learning network model, and using the first model learning module of the machine learning network model to generate an adaptive optimization instruction set corresponding to the target servo instruction set, and the first model learning module of the machine learning network model is used to use optimization to supplement the missing instruction element data in the target servo instruction set.

[0019] Continuing with the above-mentioned precision parts processing scenario of the industrial automation production line, the previously acquired target servo instruction set and the prior servo instruction set can be loaded into a specially designed machine learning network model. This machine learning network model is pre-trained, and its first model learning module has the function of optimizing the servo instructions.

[0020] In this scenario, the target servo instruction set may have some missing instruction element data due to some special requirements or limitations in the initial design. For example, when processing precision parts, due to new processing requirements, more precise control of vibration during the processing is required, but the target servo instruction set does not contain instructions specifically for such vibration control. The prior servo instruction set may contain some instruction elements related to vibration control, which may not fully meet the current high-precision requirements but can be used as a reference.

[0021] When these two instruction sets are loaded into the first model learning module of the machine learning network model, the model will analyze it based on its internal algorithm and pre-learned knowledge. For example, the model may identify the instruction pattern in the prior servo instruction set about fine-tuning the motor speed to reduce vibration in similar processing. Then, this instruction pattern is optimized in combination with the current processing requirements of the target servo instruction set, such as higher accuracy and smaller tolerance range.

[0022] Specifically, the model may analyze the command elements related to vibration control in the prior servo command set and extract the key parameters related to speed fine-tuning, such as the speed adjustment range, adjustment time interval, etc. Then, based on other command elements that already exist in the target servo command set, such as the tool feed speed, processing depth, etc., a speed fine-tuning strategy that best suits the current processing scenario is calculated. This speed fine-tuning strategy becomes part of the adaptive optimization command set to supplement the missing command element data of the target servo command set.

[0023] In addition to obtaining useful instruction elements from the prior servo instruction set, the first model learning module may also use its internal knowledge and algorithms to optimize the target servo instruction set as a whole. For example, when comprehensively analyzing instructions such as motor speed, rotation direction, and torque output, it was found that following the current instruction sequence may lead to low efficiency or fluctuations in accuracy during the processing. Therefore, the model will readjust the order of these instructions or adjust the parameter values ​​of certain instructions based on some optimization algorithms. The adaptive optimization instruction set finally generated not only includes the supplement of the missing instruction element data in the target servo instruction set, but also optimizes the entire instruction set to better adapt to the target servo work application scenario of precision parts processing.

[0024] Step S130, loading the adaptive optimization instruction set into the machine learning network model, and using the second model learning module of the machine learning network model to generate a servo system parameter configuration result of the adaptive optimization instruction set, wherein the servo system parameter information contained in the servo system parameter configuration result matches the target servo working application scenario.

[0025] In the scenario of precision parts processing in industrial automation production lines, an adaptive optimization instruction set has been obtained. Now this instruction set is loaded into the second model learning module of the machine learning network model.

[0026] This adaptive optimization instruction set contains a series of control instructions for the servo motor, such as optimized speed, torque, etc. The task of the second model learning module is to generate servo system parameter configuration results suitable for this target servo work application scenario based on these instructions.

[0027] For example, in this processing scenario, the servo system includes servo motors, drivers, controllers, and related sensors. The second model learning module will determine the optimal operating frequency and voltage parameters of the servo motor based on the speed command in the adaptive optimization instruction set. If the instruction set requires the motor to run at a higher speed during a certain processing stage, the model will calculate the voltage value that can meet this speed requirement and ensure stable operation of the motor based on the motor's performance curve and the actual processing load.

[0028] For torque commands, the model determines the appropriate torque limit that the driver should output based on the cutting force requirements between the tool and the component during the machining process. For example, when machining hard materials, a larger cutting force is required. The model calculates the driver parameter settings based on the torque settings in the adaptive optimization instruction set, the type of tool, the machining depth and other factors to ensure that sufficient torque can be provided while avoiding motor overload or affecting machining accuracy.

[0029] In addition, the model also takes into account the sensor feedback information in the entire servo system. During the processing, the sensor monitors the temperature, vibration and other parameters of the motor in real time. The second model learning module adjusts the controller parameters based on the adaptive optimization instruction set and the sensor feedback information. For example, if the sensor detects that the motor temperature is too high, the model will adjust the controller's cooling strategy parameters based on the processing progress and remaining processing tasks in the adaptive optimization instruction set, such as increasing the speed of the cooling fan or adjusting the flow of the cooling system.

[0030] The final generated servo system parameter configuration result contains the parameter information of all these servo system-related devices. These parameter information fully matches the target servo work application scenario of precision parts processing, and can ensure that the entire servo system achieves the best performance and processing accuracy when executing the adaptive optimization instruction set.

[0031] Step S140: Adaptively adjust and optimize the servo system parameters based on the servo system parameter configuration result of the adaptive optimization instruction set.

[0032] In the precision parts processing scenario of an industrial automation production line, based on the servo system parameter configuration results of the adaptive optimization instruction set obtained in the previous steps, the servo system is actually adaptively adjusted and optimized.

[0033] Each component in the servo system, such as the servo motor, drive, controller, etc., needs to be adjusted according to this parameter configuration result. For the servo motor, set the voltage and frequency parameters according to the parameter configuration result to ensure that the motor can run stably at the speed specified in the adaptive optimization instruction set. For example, if the parameter configuration result stipulates that the working voltage of the motor at a certain processing stage is 220V and the frequency is 50Hz, then set the power supply of the motor to the corresponding value so that the motor can output the appropriate speed and torque to meet the processing requirements.

[0034] As for the driver, the internal current limit and amplification factor are adjusted according to the torque limit in the servo system parameter configuration result. For example, if the torque limit is 10N·m, the driver will adjust its own circuit parameters to ensure that the torque output to the motor will not exceed this limit under any circumstances, thereby protecting the motor and processing equipment, while meeting the cutting force requirements during the processing.

[0035] The controller also needs to be adjusted accordingly. Optimize the parameters such as sensor feedback processing and processing task scheduling in the parameter configuration results. For example, if the parameter configuration results stipulate that when the motor temperature exceeds 60°C, the processing speed should be adjusted or heat dissipation measures should be added, the controller will set the corresponding temperature threshold detection and control logic. At the same time, for the scheduling of processing tasks, such as the switching time between different processing steps and the switching order of different tools, the controller will adjust according to the settings in the parameter configuration results to ensure the efficiency and accuracy of the entire processing process.

[0036] Through the adaptive parameter adjustment and optimization of the servo system parameter configuration results based on the adaptive optimization instruction set, the entire servo system can better adapt to the target servo work application scenario of precision parts processing, improve processing accuracy, reduce processing errors, improve production efficiency and extend the service life of the equipment.

[0037] Based on the above steps, the embodiment of the present application realizes efficient and accurate optimization of the servo system parameters by introducing a machine learning network model. Specifically, the method first uses the target servo instruction set and its corresponding prior servo instruction set, and generates an adaptive optimization instruction set that supplements the missing instruction element data through the first model learning module of the machine learning network model, thereby improving the integrity and accuracy of the instruction set. Furthermore, through the second model learning module of the machine learning network model, a servo system parameter configuration result that is highly matched with the target servo working application scenario is generated according to the adaptive optimization instruction set. Finally, based on the parameter configuration result, the servo system is adaptively adjusted and optimized, which significantly improves the performance and stability of the servo system in different working scenarios, realizes the intelligent and automated adjustment of the servo system parameters, and greatly improves the operating efficiency and accuracy of the servo system.

[0038] In a possible implementation, step S120 includes:

[0039] Step S121, integrate the basic network model architecture and the first special training module parameters in the machine learning network model to generate an instruction optimization sub-model, and the instruction optimization sub-model corresponds to the first model learning module of the machine learning network model.

[0040] Step S122, loading the target servo instruction set, the prior servo instruction set and the first training guidance information into the instruction optimization sub-model, using the instruction optimization sub-model to generate the adaptive optimization instruction set corresponding to the target servo instruction set, the first training guidance information is used to control the instruction optimization sub-model to optimize the target servo instruction set.

[0041] Step S130 includes:

[0042] Step S131, integrating the basic network model architecture and the second special training module parameters in the machine learning network model to generate a parameter configuration sub-model, and the parameter configuration sub-model corresponds to the second model learning module of the machine learning network model.

[0043] Step S132, load the adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model, and use the parameter configuration sub-model to generate the servo system parameter configuration result of the adaptive optimization instruction set, and the second training guidance information is used to control the parameter configuration sub-model to construct the servo system parameter information from the adaptive optimization instruction set.

[0044] In this embodiment, first, the basic network model architecture and the first special training module parameters in the machine learning network model are integrated to generate an instruction optimization sub-model. The basic network model architecture is pre-built and contains the basic structure of the neural network, such as the multi-layer neuron connection method, the type of activation function, etc. The first special training module parameters are learned based on a large amount of previous training data for the specific task of servo instruction optimization. These parameters include the relationship weights between instructions in different processing scenarios, the sensitivity to different instruction elements, etc. By integrating the two, an instruction optimization sub-model is generated, which is a model module specifically used to optimize the target servo instruction set, corresponding to the first model learning module of the machine learning network model.

[0045] Next, the target servo instruction set, the priori servo instruction set and the first training guidance information are loaded into the instruction optimization sub-model. The target servo instruction set contains various control instructions for the servo motor in the current precision parts processing scenario, such as motor speed setting, rotation direction control, torque output limit and other instructions. The priori servo instruction set is an instruction set accumulated from previous similar processing tasks, which contains some instruction modes and experience that can be learned from. The first training guidance information plays an important role in controlling the instruction optimization sub-model to optimize the target servo instruction set. For example, in this precision parts processing scenario, the first training guidance information may contain the priority setting for instruction optimization under the current processing accuracy requirements, such as giving priority to optimizing instructions related to processing efficiency while ensuring processing accuracy. The instruction optimization sub-model optimizes the target servo instruction set based on these input information, using its internal algorithms and integrated parameters. It analyzes the instruction elements in the target servo instruction set, compares them with similar instructions in the priori servo instruction set, and supplements the missing instruction element data in the target servo instruction set in combination with the optimization direction set in the first training guidance information. For example, if the target servo instruction set does not contain clear instructions for fine-tuning the motor speed at a certain processing stage, but the prior servo instruction set contains speed adjustment instructions for similar situations, and the first training guidance information indicates the importance of optimizing instructions related to processing efficiency, then the instruction optimization sub-model will generate a reasonable speed fine-tuning instruction based on these situations and supplement it to the target servo instruction set, thereby obtaining an adaptive optimization instruction set corresponding to the target servo instruction set.

[0046] For the process of loading the adaptive optimization instruction set into the machine learning network model and generating the servo system parameter configuration results using the second model learning module.

[0047] First, the basic network model architecture and the second special training module parameters in the machine learning network model are integrated to generate a parameter configuration sub-model. The basic network model architecture is the same as before, which is the basic framework of the neural network. The parameters of the second special training module are learned through a large amount of servo system related data for the task of building servo system parameter information from optimization instructions. These parameters contain the parameter mapping relationship of different types of servo system components under different instruction requirements, such as the relationship between motor voltage and frequency corresponding to different speed instructions. The parameter configuration sub-model obtained by integrating the two is specifically used to generate servo system parameter configuration results that match the adaptive optimization instruction set, corresponding to the second model learning module of the machine learning network model.

[0048] Then, the adaptive optimization instruction set and the second training guidance information are loaded into the parameter configuration submodel. The adaptive optimization instruction set contains the control instructions for the servo motor and other devices after optimization in the previous steps. The second training guidance information is used to control the parameter configuration submodel to construct the servo system parameter information from the adaptive optimization instruction set. In the precision parts processing scenario, the second training guidance information may contain the importance ranking of the components of the servo system under the current processing task and the constraints of the parameter configuration. For example, when processing high-precision parts, the motor stability during the processing is required to be the highest, so the strict requirements for the configuration of motor-related parameters will be reflected in the second training guidance information. The parameter configuration submodel analyzes the instructions in the instruction set based on the adaptive optimization instruction set and the second training guidance information. For the motor speed instruction, combined with the motor stability requirements in the second training guidance information, according to the internal integrated parameter relationship, the appropriate motor voltage, frequency and other parameters are calculated. For the torque instruction, according to the cutting force requirements during the processing and the constraints in the second training guidance information, the torque limit and other parameters of the driver are determined. At the same time, taking into account the coordination of the entire servo system, the parameters of other components such as the controller are reasonably configured according to the relationship between different instructions in the adaptive optimization instruction set and the component importance ranking in the second training guidance information, and finally the servo system parameter configuration result of the adaptive optimization instruction set is generated.

[0049] In a possible implementation, step S132 includes:

[0050] The adaptive optimization instruction set and the second training guidance information corresponding to the target servo work application scenario are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration result of the adaptive optimization instruction set, wherein different servo work application scenarios correspond to different training guidance information.

[0051] Or, the adaptive optimization instruction set and the second training guidance information corresponding to multiple servo work application scenarios are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration result of the adaptive optimization instruction set, wherein the second training guidance information is shared training guidance information of the multiple servo work application scenarios, and the target servo work application scenario belongs to the multiple servo work application scenarios.

[0052] In this embodiment, there are two situations in the above process.

[0053] In the first case, the adaptive optimization instruction set and the second training guidance information corresponding to the target servo work application scenario are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration result of the adaptive optimization instruction set, wherein different servo work application scenarios correspond to different training guidance information.

[0054] In the precision parts processing scenario, the adaptive optimization instruction set includes instructions for precise control of the servo motor's speed, torque, rotation direction, etc. The second training guidance information corresponding to the target servo work application scenario is specially customized for this precision parts processing scenario. This second training guidance information contains a lot of key content. For example, for this scenario with extremely high processing accuracy requirements, it stipulates that when configuring the servo system parameters, the parameter weight for motor speed stability is very high, which means that when configuring the servo system parameters, it is necessary to prioritize ensuring the stability of the motor speed to achieve high-precision processing. At the same time, for torque control, since the precision parts being processed have special cutting force requirements for the material, the second training guidance information will give the range limit of the torque parameter configuration and the adjustment strategy related to the processing progress.

[0055] When the adaptive optimization instruction set and this specific second training guidance information are loaded into the parameter configuration submodel, the parameter configuration submodel begins to process according to its internal algorithm and pre-learned parameter relationships. For the motor speed instruction in the adaptive optimization instruction set, the parameter configuration submodel calculates the precise voltage and frequency parameters based on the high weight requirement for speed stability in the second training guidance information and the performance parameters of the servo motor itself, such as the rated speed of the motor, the speed adjustment range, etc. During the calculation process, it will take into account the speed requirements at different stages of the entire processing process. For example, in the rough processing stage, the speed may be allowed to have a certain fluctuation range, but in the fine processing stage, the speed fluctuation must be strictly controlled within a very small range in accordance with the requirements in the second training guidance information, so as to determine the voltage and frequency parameters corresponding to different stages to ensure the stability of the motor speed.

[0056] For the torque command, the torque output parameters of the driver are determined according to the torque parameter configuration range limit in the second training guidance information and the adjustment strategy related to the processing progress, combined with factors such as the type of tool and processing depth. If during the processing, as the processing depth increases, a greater cutting force is required, the parameter configuration sub-model will gradually adjust the output torque of the driver to meet the processing requirements based on the adjustment strategy in the second training guidance information, without exceeding the torque parameter configuration range limit. At the same time, considering the collaborative work of the entire servo system, the parameter configuration sub-model will also reasonably configure the parameters of other servo system components such as the controller, such as the feedback adjustment parameters of the controller, the communication protocol parameters, etc., according to the relationship between other instructions and the speed and torque instructions in the adaptive optimization instruction set, and the requirements for the stability of the entire system in the second training guidance information, and finally generate a servo system parameter configuration result that meets this precision parts processing scenario.

[0057] In the second case, the adaptive optimization instruction set and the second training guidance information corresponding to multiple servo work application scenarios are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration results of the adaptive optimization instruction set, wherein the second training guidance information is shared training guidance information for multiple servo work application scenarios, and the target servo work application scenario belongs to multiple servo work application scenarios.

[0058] In an industrial automation production line, it is assumed that there are multiple servo work application scenarios. In addition to precision parts processing scenarios, there are also general parts processing scenarios and parts assembly scenarios. These scenarios have some commonalities, so there can be shared second training guidance information. This shared second training guidance information covers some common requirements and constraints for multiple scenarios. For example, for the energy efficiency requirements of the entire production line, this needs to be considered in each scenario, so the shared second training guidance information will include guidelines for optimizing the energy consumption of the servo system, such as minimizing the energy consumption of the motor while meeting the processing or assembly requirements.

[0059] When the adaptive optimization instruction set and this shared second training guidance information are loaded into the parameter configuration sub-model, the parameter configuration sub-model will process it according to the energy consumption optimization principle in the shared second training guidance information. For the motor speed instructions in the adaptive optimization instruction set, the parameter configuration sub-model will consider how to reduce energy consumption by adjusting the motor speed while ensuring normal processing or assembly in different scenarios. In the precision parts processing scenario, although processing accuracy is the primary consideration, energy consumption optimization cannot be ignored. The parameter configuration sub-model will fine-tune the motor speed based on the energy consumption optimization principle in the shared second training guidance information, combined with the specific requirements of precision parts processing, such as meeting the processing accuracy, to find a speed value that can both ensure accuracy and reduce energy consumption, and then determine the corresponding voltage and frequency parameters.

[0060] For torque instructions, under the guidance of the shared second training guidance information, the parameter configuration sub-model will consider the commonality of torque requirements in different scenarios. In the precision parts processing scenario, although the torque demand has its particularity, it is also necessary to follow the shared energy consumption optimization principle. For example, when determining the torque output parameters of the driver, the energy consumption of the entire production line will be comprehensively considered to avoid excessive output torque causing energy waste. At the same time, according to the relationship between other instructions and speed and torque instructions in the adaptive optimization instruction set, as well as the requirements for the stability and versatility of the entire servo system in the shared second training guidance information, the parameters of other servo system components such as the controller are reasonably configured, so that the generated servo system parameter configuration results not only meet the special requirements of the target servo work application scenario of precision parts processing, but also meet some basic principles shared by multiple servo work application scenarios, such as energy consumption optimization and system stability.

[0061] In a possible implementation, the method further includes:

[0062] Step A110, obtaining template parameter configuration learning data for different servo work application scenarios, wherein the template parameter configuration learning data includes a template adaptive optimization instruction set and annotated servo system parameter information corresponding to the template adaptive optimization instruction set.

[0063] Step A120, loading the template adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model, and using the parameter configuration sub-model to generate a template servo system parameter configuration result of the template adaptive optimization instruction set.

[0064] Step A130: determining a parameter configuration error according to the template servo system parameter configuration result and the annotated servo system parameter information.

[0065] Step A140: Optimize the second special training module parameters and the second training guidance information according to the parameter configuration error.

[0066] In a possible implementation, step A120 includes: loading the template adaptive optimization instruction set and the second training guidance information corresponding to the servo work application scenario to which the template adaptive optimization instruction set belongs into the parameter configuration sub-model, and using the parameter configuration sub-model to generate the template servo system parameter configuration result of the template adaptive optimization instruction set.

[0067] Step A140 includes: optimizing the second special training module parameters and the second training guidance information corresponding to different servo work application scenarios according to the parameter configuration error.

[0068] In this embodiment, first, template parameter configuration learning data for different servo work application scenarios are obtained. Under the general framework of industrial automation, there are a variety of servo work application scenarios, such as the precision parts processing scenario mentioned above, general parts processing scenario, and parts assembly scenario. For each scenario, there is corresponding template parameter configuration learning data. The template adaptive optimization instruction set in these data contains instruction information optimized for the servo system in a specific scenario. Taking the precision parts processing scenario as an example, the template adaptive optimization instruction set may contain instructions for accurately controlling the servo motor speed, torque, rotation direction, etc. under a specific processing technology. These instructions are preliminarily optimized to meet the needs of precision parts processing. At the same time, the marked servo system parameter information corresponding to the template adaptive optimization instruction set is a predetermined accurate servo system parameter configuration. These parameters are obtained through actual verification or based on accurate theoretical calculations, which can ensure that the servo system reaches the best operating state in this scenario. For example, for a servo motor, the marked servo system parameter information clearly specifies the optimal voltage, frequency, torque limit and other parameters at different processing stages. These parameters are set to achieve the goals of high precision and high efficiency required for precision parts processing.

[0069] Next, the template adaptive optimization instruction set and the second training guidance information are loaded into the parameter configuration sub-model. The parameter configuration sub-model here is used to generate the servo system parameter configuration results based on the input instruction set and training guidance information. In the precision parts processing scenario, the template adaptive optimization instruction set is loaded, and the various types of instruction information contained therein become important inputs to the parameter configuration sub-model. The second training guidance information is specific information corresponding to the scenario, which contains the guiding principles for the servo system parameter configuration in this scenario. For example, in the precision parts processing scenario, due to the extremely high processing accuracy requirements, the second training guidance information may emphasize the importance of motor speed stability during the parameter configuration process, and also has strict requirements on the accuracy and stability of torque output. After this embodiment loads this information into the parameter configuration sub-model, the parameter configuration sub-model performs calculations based on its internal algorithm and pre-learned parameter relationships. For the motor speed command in the template adaptive optimization instruction set, the parameter configuration sub-model calculates the motor speed related parameters in the template servo system parameter configuration result in this scenario, such as the motor's operating voltage, frequency, etc., based on the requirements for speed stability in the second training guidance information, combined with the instruction content of the template adaptive optimization instruction set itself and the performance parameters of the servo motor. For the torque command, the torque output parameters of the driver are determined according to the requirements in the second training guidance information and the content of the template adaptive optimization instruction set. At the same time, considering the synergy of the entire servo system, the parameters of other components such as the controller will also be calculated and determined accordingly, so as to generate a complete template servo system parameter configuration result.

[0070] Then, the parameter configuration error is determined based on the template servo system parameter configuration result and the annotated servo system parameter information. In the precision parts processing scenario, the generated template servo system parameter configuration result is compared with the annotated servo system parameter information. For example, for the motor speed related parameters, the motor operating voltage and frequency in the template servo system parameter configuration result are compared with the standard values ​​in the annotated servo system parameter information. If the motor operating voltage in the template servo system parameter configuration result is 220V, and the standard voltage in the annotated servo system parameter information is 225V, this will produce a certain difference. Similarly, such a comparison will be made for torque output parameters, controller related parameters, etc. The combination of these differences reflects the parameter configuration error. This error is not just a simple numerical difference, but also involves an evaluation of the impact on the performance of the entire servo system. For example, a small difference in motor voltage may cause fluctuations in motor speed, which in turn affects processing accuracy or processing efficiency. These effects must be comprehensively considered when determining the parameter configuration error.

[0071] Finally, the second special training module parameters and the second training guidance information are optimized according to the parameter configuration error. Since the parameter configuration error reflects the degree of deviation of the current parameter configuration sub-model from the ideal state (marking the servo system parameter information) when generating the servo system parameter configuration result, it is necessary to optimize the relevant parameters and information. In the processing logic of this embodiment, for the second special training module parameters, these parameters are closely related to the training and operation of the parameter configuration sub-model. For example, in the precision parts processing scenario, if it is found that the parameter configuration error related to the motor speed is large, the part related to the motor speed calculation in the second special training module parameters can be analyzed, and it may be that some weight parameters or mapping relationships need to be adjusted. This embodiment can optimize these parameters according to the specific situation of the error to improve the accuracy of the parameter configuration sub-model in calculating the motor speed related parameters. At the same time, for the second training guidance information, since the second training guidance information corresponding to different servo work application scenarios also needs to be optimized, in the precision parts processing scenario, if it is found that the parameter configuration error is caused by the unreasonable weight setting of the motor speed stability in the second training guidance information, this part of the content can be adjusted. For other scenarios, such as general parts processing scenarios or parts assembly scenarios, although each has its own characteristics, the corresponding second special training module parameters and second training guidance information will be optimized according to the overall parameter configuration error to improve the parameter configuration accuracy of the entire servo system in different application scenarios and the performance of the servo system. Through such an optimization process, the performance of the parameter configuration sub-model can be continuously improved, so that it can generate more accurate and more practical results in the subsequent servo system parameter configuration process.

[0072] In a possible implementation, the method further includes:

[0073] Step B110, based on the correlation between the servo work application scenario and the instruction elements in the instruction knowledge set, determine x reference instruction elements from the instruction knowledge set, the correlation between the reference instruction elements and the servo work application scenario is greater than the correlation between other instruction elements and the servo work application scenario, and the instruction knowledge set is obtained by iterative learning during the model parameter optimization process of the basic network model architecture.

[0074] Step B120: Initialize the second training guidance information corresponding to the servo work application scenario according to the x reference instruction elements and the random instruction elements.

[0075] In a possible implementation, step B120 includes:

[0076] Step B121, performing attribute analysis on the input x reference instruction elements, and dividing the x reference instruction elements into different attribute categories according to the attribute analysis result.

[0077] Step B122, analyze the position, functional tendency and potential association with other known instruction elements of the random instruction element in the entire instruction system to determine the attribute supplement direction information of the random instruction element in the training guidance information, and the attribute supplement direction information is used to guide the fusion process of the random instruction element and the reference instruction element.

[0078] Step B123, constructing a preliminary framework structure according to the attribute categories of the reference instruction elements, wherein the preliminary framework structure is used to express the relationship between different attribute categories and the logical order of different attribute categories in the servo work application scenario.

[0079] Step B124, supplementing direction information according to the attributes of the random instruction element, reserving a position for integrating the random instruction element in the preliminary framework structure, and generating an intermediate framework structure including the classification relationship of the reference instruction element and the integration position of the random instruction element.

[0080] Step B125, determines the rules for setting the weights of the instruction elements based on the importance of different parts in the intermediate framework structure and the degree of influence in the servo work application scenario, wherein, for random instruction elements, the range and adjustment direction of the initial weight setting of the random instruction elements are determined based on the attribute supplementary direction information and the integration position in the intermediate framework structure.

[0081] Step B126, according to the rule for setting the weight of the instruction element, weights are allocated to the reference instruction element and the random instruction element, and an instruction element set including all the reference instruction elements and random instruction elements and with weight information is generated.

[0082] Step B127, based on the set of instruction elements with weights, construct an instruction element association matrix, wherein the rows and columns of the instruction element association matrix are instruction elements, and the elements in the instruction element association matrix represent the degree of association between two instruction elements, and the degree of association is determined based on the weights of the instruction elements, the attribute categories and the logical relationships in the framework structure.

[0083] Step B128, converting the degree of association in the instruction element association matrix into the priority order and influencing factor of different instruction elements when constructing parameter information, and at the same time determining the importance of different instruction elements in the training guidance information based on the weight information of the instruction elements, and constructing and generating the second training guidance information.

[0084] In this embodiment, first, according to the correlation between the servo work application scenario and the instruction elements in the instruction knowledge set, x reference instruction elements are determined from the instruction knowledge set. The instruction knowledge set is obtained by iterative learning in the process of optimizing the model parameters of the basic network model architecture, and it contains many instruction elements related to the servo system. In the servo work application scenario of precision parts processing, the instruction knowledge set covers the instruction-related knowledge accumulated from different processing tasks, different equipment configurations and various operating conditions. The server determines x reference instruction elements by analyzing the correlation between each instruction element and the precision parts processing scenario. For example, for precision parts with extremely high processing accuracy requirements, instruction elements related to precise control of motor speed, fine-tuning instruction elements of tool feed amount, and related instruction elements for vibration suppression during processing, etc., are closely related to the realization of precision processing, so they will be determined as reference instruction elements. Compared with other instruction elements, such as some instruction elements related to the equipment initialization process but with little effect on processing accuracy, these reference instruction elements are more closely related to the precision parts processing scenario.

[0085] Next, the second training guidance information corresponding to the servo work application scenario is initialized according to the x reference instruction elements and the random instruction elements.

[0086] In this process, the attributes of the x input reference instruction elements are first analyzed, and then the x reference instruction elements are divided into different attribute categories based on the attribute analysis results. In the precision parts processing scenario, the attributes of the reference instruction elements related to the precise control of the motor speed may be classified as the power control attribute category, because the motor speed directly affects the power output during the processing. The reference instruction elements related to the fine-tuning of the tool feed rate may be classified as the processing technology control attribute category, because the tool feed rate is a key factor in the processing process. The instruction elements related to vibration suppression during the processing can be classified as the processing stability control attribute category.

[0087] Then, the position, functional tendency and potential association with other known instruction elements of the random instruction element in the entire instruction system are analyzed to determine the attribute supplement direction information of the random instruction element in the supplement to the training guidance information. Suppose there is a random instruction element, which is related to the overload protection mechanism of the motor in the entire instruction system. In the precision parts processing scenario, the stable operation of the motor is crucial to the processing accuracy, so the functional tendency of this random instruction element is to ensure the safe and stable operation of the motor and prevent the processing accuracy from being affected by overload. It has a potential association with the motor speed control instruction element, because excessive motor speed may cause overload. Then its attribute supplement direction information supplemented to the training guidance information is related to the power control attribute category, and focuses on supplementing the power control from the perspective of protecting the motor.

[0088] According to the attribute categories of the reference instruction elements, a preliminary framework structure is constructed. In the precision parts processing scenario, this preliminary framework structure reflects the relationship between different attribute categories and the logical order in the scenario. For example, the power control attribute category may be the first to be considered, because the normal operation of the motor is the basis of processing, followed by the processing technology control attribute category, because it determines the specific processing operation mode, and finally the processing stability control attribute category, which is the guarantee to ensure the processing accuracy during the processing. This preliminary framework structure links different attribute categories in a logical order, expressing the order and relationship of each attribute category in the entire processing process in the precision parts processing scenario.

[0089] According to the attributes of the random instruction elements, the direction information is supplemented, and a position for the integration of the random instruction elements is reserved in the preliminary framework structure, generating an intermediate framework structure containing the classification relationship of the reference instruction elements and the position of the random instruction elements. For the random instruction elements related to motor overload protection mentioned above, the direction information is supplemented according to their attributes, and a position is reserved under the power control attribute category. This position indicates its relative relationship with other power control related instruction elements, thereby generating an intermediate framework structure. This intermediate framework structure contains both the classification relationship of the reference instruction elements and the specific position of the random instruction elements, so that subsequent operations can be carried out in an orderly framework.

[0090] According to the importance of different parts in the intermediate framework structure and the degree of influence in the servo work application scenario, the rules for setting the weight of the instruction element are determined. In the precision parts processing scenario, the power control attribute category is the basis of processing, so the instruction elements in this attribute category may have a higher initial weight. For random instruction elements, the range and adjustment direction of their initial weight setting are determined according to their attribute supplementary direction information and the integration position in the intermediate framework structure. For example, the random instruction element related to motor overload protection is important, but its weight is slightly lower than that of core power control instruction elements such as motor speed, because it is under the power control attribute category and is a protection mechanism for the normal operation of the motor. The range of its weight setting may be a relatively small value, and as the motor load changes during the processing process, its weight has the direction of upward adjustment. For example, when the motor load is close to the overload threshold, the weight of this random instruction element will increase to further strengthen the overload protection function of the motor.

[0091] According to the rules for setting the weight of instruction elements, the reference instruction elements and random instruction elements are weighted, and a set of instruction elements with weight information is generated, which includes all reference instruction elements and random instruction elements. In the precision parts processing scenario, a higher weight, such as 0.3, is assigned to the motor speed control instruction element in the power control attribute category, while the tool feed fine-tuning instruction element may have a weight of 0.2 in the processing technology control attribute category, and the vibration suppression instruction element in the processing process has a weight of 0.15 in the processing stability control attribute category. The weight of the random instruction element related to motor overload protection is set to 0.05, etc. In this way, a set of instruction elements containing all relevant instruction elements with clear weight information is generated.

[0092] Based on the set of weighted instruction elements, an instruction element association matrix is ​​constructed. The rows and columns of this instruction element association matrix are instruction elements, and the elements represent the degree of association between two instruction elements. In the precision parts processing scenario, there is a certain degree of association between the motor speed control instruction element and the tool feed amount fine-tuning instruction element, because the motor speed will affect the tool feed speed. The determination of this degree of association is based on the weight of the instruction element, the attribute category, and the logical relationship in the framework structure. For example, the weight of the motor speed control instruction element is 0.3, and the weight of the tool feed amount fine-tuning instruction element is 0.2. They belong to the power control attribute category and the processing technology control attribute category respectively, and there is a logical order in the processing process. The degree of association between them is determined comprehensively based on these factors.

[0093] Finally, the degree of association in the instruction element association matrix is ​​converted into the priority order and influencing factor of different instruction elements when constructing parameter information. At the same time, according to the weight information of the instruction elements, the importance of different instruction elements in the training guidance information is determined to construct the second training guidance information. In the precision parts processing scenario, the instruction elements with high association have a higher priority order when constructing parameter information. For example, the motor speed control instruction element will be given priority when constructing parameter information because it has a high degree of association with other multiple instruction elements. Its influencing factor is also relatively large, which has a greater impact on the final servo system parameter configuration results. According to the weight information of the instruction elements, the motor speed control instruction element has a higher degree of importance in the training guidance information, which will guide the parameter configuration sub-model to focus on the factors related to the motor speed when generating the servo system parameter configuration results. Through such a series of steps, the second training guidance information generated can accurately reflect the relationship and importance of each instruction element in the precision parts processing scenario, thereby providing effective guidance for the subsequent servo system parameter configuration.

[0094] In a possible implementation, step A120 includes: loading the template adaptive optimization instruction set and the second training guidance information corresponding to multiple servo work application scenarios into the parameter configuration sub-model, and using the parameter configuration sub-model to generate the template servo system parameter configuration result of the template adaptive optimization instruction set.

[0095] Step A140 includes: optimizing the second special training module parameters and the second training guidance information corresponding to the multiple servo work application scenarios according to the parameter configuration error.

[0096] In this embodiment, multiple servo work application scenarios are taken into consideration. In the example of an industrial automation production line, there are multiple servo work application scenarios such as precision parts processing scenarios, general parts processing scenarios, and parts assembly scenarios.

[0097] The template adaptive optimization instruction set contains instruction optimization information for different types of processing or assembly operations. For example, in the precision parts processing scenario, the template adaptive optimization instruction set may contain special speed adjustment instructions for the servo motor at different processing stages to meet the high-precision processing requirements; in the general parts processing scenario, it may contain some speed and torque adjustment instructions that focus more on processing efficiency; in the parts assembly scenario, there may be specific position control instructions that cooperate with the assembly action. The second training guidance information corresponding to multiple servo work application scenarios is guidance information that integrates the needs of each scenario. For example, in terms of energy efficiency, since the entire production line has the requirement to reduce energy consumption, the second training guidance information will include a general strategy for optimizing the energy consumption of the servo system in different scenarios, such as reasonably adjusting the working voltage of the motor according to the motor load conditions in different scenarios under the premise of meeting the basic requirements of processing or assembly. At the same time, special needs for different scenarios will also be reflected, such as the emphasis on the configuration of parameters related to processing accuracy in the precision parts processing scenario, the consideration of the configuration of parameters related to processing speed improvement in the general parts processing scenario, and the attention to the configuration of parameters related to assembly action accuracy in the parts assembly scenario.

[0098] When the template adaptive optimization instruction set and the second training guidance information corresponding to these multiple servo working application scenarios are loaded into the parameter configuration sub-model, the parameter configuration sub-model processes according to its internal algorithm and predefined parameter relationship. For the motor-related instructions in the template adaptive optimization instruction set, the parameter configuration sub-model will calculate the working parameters of the motor according to the energy consumption optimization strategy in the second training guidance information and the special requirements of each scenario. In the precision parts processing scenario, if the instruction set has a high speed requirement for a certain processing stage, the parameter configuration sub-model will determine the appropriate voltage and frequency parameters in combination with the energy consumption optimization strategy while meeting the processing accuracy requirements (according to the precision processing related configuration requirements in the second training guidance information). In the general parts processing scenario, for the same speed requirement, it may be more inclined to choose a parameter configuration that can increase the processing speed while taking into account a certain energy consumption optimization. For torque instructions, in the parts assembly scenario, according to the force requirements of the assembly action (from the template adaptive optimization instruction set) and the requirements for the accuracy of the assembly action in the second training guidance information, the torque parameters output by the driver are determined to ensure the correct assembly of the parts during the assembly process without causing damage to the parts. In this way, the parameter configuration sub-model generates a template servo system parameter configuration result for the template adaptive optimization instruction set that fully considers the requirements of multiple servo work application scenarios.

[0099] Next, the parameters of the second special training module and the second training guidance information corresponding to multiple servo work application scenarios are optimized based on the parameter configuration error. In the above-mentioned multi-scenario industrial automation production line, the parameter configuration error reflects the degree of deviation between the template servo system parameter configuration results generated by the parameter configuration sub-model and the ideal state. For example, in the precision parts processing scenario, if the motor speed related parameters in the template servo system parameter configuration results deviate from the actual optimal parameters, this will produce a parameter configuration error. This error may cause the processing accuracy to fail to meet the standard or reduce the processing efficiency. Similar situations also exist for general parts processing scenarios and parts assembly scenarios. For example, in general parts processing scenarios, if the torque parameter configuration is unreasonable, it may affect the processing speed or processing quality; in the parts assembly scenario, errors in position control related parameters may cause assembly failure or reduced assembly accuracy.

[0100] In order to optimize this situation, it is necessary to adjust the parameters of the second special training module and the second training guidance information corresponding to multiple servo work application scenarios based on these parameter configuration errors. For the second special training module parameters, these parameters are closely related to the processing capabilities of the parameter configuration sub-model for different instructions and scenarios. If it is found that the parameter configuration error related to the motor speed is large in the precision parts processing scenario, it may be necessary to adjust the part of the second special training module parameters related to the calculation and adjustment of the motor speed, such as adjusting certain parameters related to the motor performance curve fitting, or parameters related to the speed switching strategy in different processing stages. The second training guidance information corresponding to multiple servo work application scenarios also needs to be optimized. For example, when it is found that the energy consumption optimization effect in multiple scenarios does not meet expectations, it may be necessary to review the energy consumption optimization strategy in the second training guidance information and adjust the trade-off relationship between parameters such as motor voltage and torque and energy consumption in different scenarios. In the precision parts processing scenario, if the parameter configuration error related to the processing accuracy is shown to be caused by unreasonable weight setting of the accuracy requirement in the second training guidance information, it is necessary to adjust this part of the weight; in the general parts processing scenario, if the error related to the processing speed shows that the attention to improving the processing speed is insufficiently reflected in the second training guidance information, the corresponding content also needs to be optimized; in the parts assembly scenario, the corresponding part of the second training guidance information should be improved according to the error related to the accuracy of the assembly action.

[0101] In a possible implementation, the method further includes:

[0102] Step C110, based on the correlation between the servo work application scenario and the instruction elements in the instruction knowledge set, determine x reference instruction elements from the instruction knowledge set, the correlation between the reference instruction elements and the servo work application scenario is greater than the correlation between other instruction elements and the servo work application scenario, and the instruction knowledge set is obtained by iterative learning during the model parameter optimization process of the basic network model architecture.

[0103] Step C120 , fusing the x reference instruction elements corresponding to different servo work application scenarios to generate fused instruction elements corresponding to the multiple servo work application scenarios.

[0104] Step C130 , initializing the second training guidance information corresponding to the plurality of servo work application scenarios according to the fusion instruction element and the random instruction element.

[0105] In this embodiment, the instruction knowledge set is obtained by iterative learning during the model parameter optimization process of the basic network model architecture, which contains a large number of instruction elements related to the operation of the servo system. Taking the industrial automation production line as an example, in the precision parts processing scenario, the instruction elements closely related to the processing accuracy, such as the precise speed control instruction of the motor under small displacement, the feed control instruction of the tool during ultra-fine processing, etc., are determined as reference instruction elements because they are critical to the high-quality completion of precision parts processing. In the general parts processing scenario, those instruction elements related to fast processing, such as the motor efficient operation instruction under large torque, the instruction related to fast tool switching, etc., will be selected as reference instruction elements because of their strong correlation with improving the processing speed. In the parts assembly scenario, the instruction elements related to the precise position control of the assembly parts, such as the precise control instruction of the servo motor under small angle rotation, etc., are determined as reference instruction elements because they are crucial to the accuracy of the assembly action.

[0106] Then, the x reference instruction elements corresponding to different servo working application scenarios are integrated to generate fused instruction elements corresponding to multiple servo working application scenarios. Taking the industrial automation production line as an example, the reference instruction elements related to processing accuracy determined in the precision parts processing scenario, such as the motor micro-displacement speed control instruction, and the motor large torque operation instruction related to efficient processing in the general parts processing scenario and the small angle motor control instruction in the parts assembly scenario are integrated. This fusion is not a simple superposition, but an integration based on the importance of different instruction elements in their respective scenarios and the comprehensive impact on the function of the entire servo system. For example, the fused instruction element may include a new control strategy, that is, based on the large torque operation instruction in the general parts processing scenario, the precision control idea in the precision parts processing scenario is borrowed to ensure the accuracy of the large torque operation process to a certain extent, and at the same time, combined with the stability requirements of small angle control in the parts assembly scenario, a fused instruction element is formed that can not only meet the basic needs of different scenarios, but also improve the overall servo system performance to a certain extent.

[0107] Finally, the second training guidance information corresponding to multiple servo work application scenarios is initialized based on the fused instruction element and the random instruction element. In this process, the fused instruction element contains the comprehensive characteristics of the key instruction elements in multiple servo work application scenarios. The random instruction element may be some instruction elements introduced under special circumstances or to further optimize the system. For example, in an industrial automation production line, there may be a random instruction element that is about the protection instruction for the servo system in the case of sudden power fluctuations. The second training guidance information is initialized according to the different parts of the fused instruction element, such as the accuracy-related part in the precision parts processing scenario, the efficiency-related part in the general parts processing scenario, and the accuracy-related part in the parts assembly scenario, as well as the functional characteristics of the random instruction element. For the accuracy-related part in the fused instruction element, a higher weight will be set in the second training guidance information to ensure the processing accuracy in the precision parts processing scenario; for the efficiency-related part, an appropriate weight will be given in the general parts processing scenario to reflect the concern for processing speed; for the accuracy-related part, an appropriate weight is determined in the parts assembly scenario to ensure the accuracy of the assembly. For the random instruction element, since it is an instruction about power fluctuation protection, its weight and influence in the second training guidance information will be determined according to the stability of the power supply in the entire production line and the sensitivity of the servo system to power fluctuations in each scenario. For example, if the power supply is unstable and is most sensitive to power fluctuations in the precision parts processing scenario, then the weight of this random instruction element in the second training guidance information in the precision parts processing scenario will be relatively high, and the parameter settings related to power fluctuation protection will be given priority during parameter configuration to ensure that the quality of precision parts processing is not greatly affected in this case. In this way, the initialized second training guidance information can comprehensively consider the needs and special circumstances of multiple servo work application scenarios, and provide effective guidance for subsequent servo system parameter configuration.

[0108] In one possible implementation, the method includes:

[0109] Step D110, obtaining template instruction optimization learning data for different servo work application scenarios, wherein the template instruction optimization learning data includes a template servo instruction set, a template priori servo instruction set, and a labeled adaptive optimization instruction set corresponding to the template servo instruction set.

[0110] Step D120, loading the template servo instruction set, the template priori servo instruction set and the first training guidance information into the instruction optimization sub-model, and using the instruction optimization sub-model to generate a template adaptive optimization instruction set corresponding to the template servo instruction set.

[0111] Step D130, determining the instruction optimization error according to the template adaptive optimization instruction set and the label adaptive optimization instruction set.

[0112] Step D140, optimizing the first special training module parameters according to the instruction optimization error.

[0113] Wherein, step D110 includes:

[0114] Step D111, loading the template servo instruction set and the template priori servo instruction set into a first machine learning network, and using the first machine learning network to generate a first servo system parameter configuration result of the template servo instruction set.

[0115] Step D112, loading the instruction optimization sample, the template servo instruction set and the template priori servo instruction set into a second machine learning network, and using the second machine learning network to generate a reference adaptive optimization instruction set corresponding to the template servo instruction set.

[0116] Step D113, generating the template instruction optimization learning data for different servo work application scenarios based on the comparison result of the first servo system parameter configuration result and the reference adaptive optimization instruction set.

[0117] In this embodiment, there are multiple servo work application scenarios in industrial automation production lines, such as precision parts processing scenarios, general parts processing scenarios, and parts assembly scenarios, etc. For each scenario, it is necessary to obtain corresponding template instruction optimization learning data.

[0118] The template instruction optimization learning data includes the template servo instruction set, the template priori servo instruction set, and the annotated adaptive optimization instruction set corresponding to the template servo instruction set. Taking the precision parts processing scenario as an example, the template servo instruction set contains a series of instructions for the initial setting of the servo system in this scenario. These instructions cover the speed control instructions, torque control instructions, rotation direction instructions, etc. of the servo motor, which are set to realize the basic operations of precision parts processing. The template priori servo instruction set is an instruction set accumulated from previous similar precision parts processing tasks or related processing experience. It contains some instruction modes or parameter settings that have been proven to be helpful for similar processing tasks in the past. The annotated adaptive optimization instruction set corresponding to the template servo instruction set is an optimized instruction set obtained through precise calculation or under ideal processing conditions. This instruction set represents the instruction combination that can enable the servo system to achieve the best performance in the precision parts processing scenario, such as optimizing the processing efficiency while meeting the highest processing accuracy requirements.

[0119] The process of obtaining these template instruction optimization learning data is relatively complicated. First, the template servo instruction set and the template prior servo instruction set are loaded into the first machine learning network. In the precision parts processing scenario, the first machine learning network is a network that has been trained with a large amount of data, and it can generate the corresponding servo system parameter configuration results according to the input instruction set. After the template servo instruction set and the template prior servo instruction set are loaded, the first machine learning network performs calculations based on its internal algorithm and predefined model parameters. For example, for the motor speed instruction in the template servo instruction set, the first machine learning network will calculate the motor speed related parameters in a preliminary first servo system parameter configuration result based on the speed setting of the similar processing stage in the template prior servo instruction set and the performance parameters of the motor. For example, in a certain processing stage, according to the rated power, load conditions and processing accuracy requirements of the motor, the working voltage and frequency of the motor are calculated. Parameters, these parameters constitute part of the first servo system parameter configuration result of the template servo instruction set.

[0120] Next, the instruction optimization examples, template servo instruction set, and template prior servo instruction set are loaded into the second machine learning network. The instruction optimization examples are some optimized instruction examples that have been manually screened or extracted from successful processing cases. These examples contain effective instruction optimization methods for different processing situations. In the precision parts processing scenario, the second machine learning network also processes these inputs based on its own algorithms and model parameters. For example, the instruction optimization examples may include successful examples of fine-tuning the motor speed under specific processing accuracy requirements. When loaded into the second machine learning network together with the template servo instruction set and the template prior servo instruction set, the second machine learning network will generate a reference adaptive optimization instruction set corresponding to the template servo instruction set based on this information. This reference adaptive optimization instruction set is a reference to an optimization instruction set based on the instruction optimization examples and the existing template servo instruction set and template prior servo instruction set. It may contain some new speed adjustment strategies, torque optimization methods, etc., aiming to improve the performance of precision parts processing.

[0121] Finally, based on the comparison results of the first servo system parameter configuration results and the reference adaptive optimization instruction set, template instruction optimization learning data for different servo work application scenarios are generated. In the precision parts processing scenario, the first servo system parameter configuration results and the reference adaptive optimization instruction set are compared in detail. For example, by comparing the motor speed related parameters in the two, the motor operating voltage in the first servo system parameter configuration result is 220V, while the motor operating voltage corresponding to the same processing stage in the reference adaptive optimization instruction set is 215V. This difference becomes an important basis for generating template instruction optimization learning data. The same comparison is also performed for other parameters such as torque and direction of rotation. Through this comparison, it is possible to find out how to optimize the direction and method of instructions based on the template servo instruction set according to the existing prior experience and optimization examples, thereby generating template instruction optimization learning data for precision parts processing scenarios.

[0122] After obtaining the template instruction optimization learning data, the template servo instruction set, the template priori servo instruction set and the first training guidance information are loaded into the instruction optimization sub-model, and the template adaptive optimization instruction set corresponding to the template servo instruction set is generated by the instruction optimization sub-model. In the precision parts processing scenario, the template servo instruction set and the template priori servo instruction set are loaded into the instruction optimization sub-model, and the first training guidance information is used to guide the instruction optimization sub-model to optimize the template servo instruction set. For example, the first training guidance information may include the requirements for the priority guarantee of processing accuracy in the current precision parts processing scenario, and the expectation of improving processing efficiency while ensuring accuracy. The instruction optimization sub-model performs optimization calculations based on these input information. For the motor speed instruction in the template servo instruction set, the instruction optimization sub-model will generate an optimized speed instruction based on the relevant experience in the template priori servo instruction set and the accuracy and efficiency requirements in the first training guidance information, combined with the performance curve of the motor and the actual load conditions. For example, if the speed instruction in the template servo instruction set is 1000 rpm at a certain processing stage, the instruction optimization sub-model may optimize it to 1050 rpm based on prior experience and guidance information to improve processing efficiency while meeting accuracy requirements. Similarly, similar optimization is performed for torque instructions, rotation direction instructions, etc., and finally a template adaptive optimization instruction set corresponding to the template servo instruction set is generated.

[0123] Next, the instruction optimization error is determined based on the template adaptive optimization instruction set and the labeled adaptive optimization instruction set. In the precision parts processing scenario, the generated template adaptive optimization instruction set is comprehensively compared with the labeled adaptive optimization instruction set. For example, for the motor speed instruction, the speed of a certain processing stage in the template adaptive optimization instruction set is 1050 rpm, while the speed of the labeled adaptive optimization instruction set in the same processing stage is 1030 rpm, which results in an error in the speed instruction. Similar comparisons and calculations are also performed for other instructions such as torque instructions and rotation direction instructions, and the instruction optimization error is obtained by combining these differences. This instruction optimization error reflects the degree of deviation of the instruction optimization sub-model from the ideal labeled adaptive optimization instruction set when generating the template adaptive optimization instruction set. It is not just a simple numerical difference, but also involves an evaluation of the impact on the servo system performance during the entire precision parts processing process.

[0124] Finally, the parameters of the first special training module are optimized according to the instruction optimization error. In the precision parts processing scenario, due to the existence of instruction optimization errors, it is indicated that the parameters of the first special training module need to be optimized. For example, if the error in the motor speed instruction is large, it may mean that there are problems in the first special training module parameters related to the motor speed optimization calculation, such as some weight parameters are set unreasonably or the fitting relationship with the motor performance curve is inaccurate. The server will adjust these related first special training module parameters according to the specific situation of the instruction optimization error. For example, adjust the weight parameters related to the motor speed calculation, or refit the relationship between the motor performance curve and the instruction optimization, etc. Through such an optimization process, the accuracy and effectiveness of the instruction optimization sub-model in processing template servo instruction sets in similar precision parts processing scenarios are improved, so that it can generate results that are closer to the labeled adaptive optimization instruction set, thereby improving the performance of the entire servo system in the precision parts processing scenario.

[0125] Figure 2 FIG. 1 shows the hardware structure of a servo system adaptive parameter adjustment optimization system 100 based on machine learning for implementing the above-mentioned servo system adaptive parameter adjustment optimization method based on machine learning provided by an embodiment of the present invention, as shown in FIG. Figure 2 As shown, the servo system adaptive parameter adjustment optimization system 100 based on machine learning may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0126] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the machine-learning-based servo system adaptive parameter adjustment optimization system 100 uses to execute or use to complete the exemplary method described in the present invention.

[0127] In the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the servo system adaptive parameter adjustment optimization method based on machine learning in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0128] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned servo system adaptive parameter adjustment and optimization system 100 based on machine learning. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0129] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above-mentioned servo system adaptive parameter adjustment optimization method based on machine learning is implemented.

[0130] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A servo system adaptive parameter adjustment optimization method based on machine learning, characterized in that: The method comprises: Acquire a target servo instruction set of a servo system, and a priori servo instruction set corresponding to the target servo instruction set, wherein the target servo instruction set and the priori servo instruction set are servo instruction sets for target servo working application scenarios, and the priori servo instruction set and the target servo instruction set have instruction association features; The target servo instruction set and the priori servo instruction set are loaded into a machine learning network model, and an adaptive optimization instruction set corresponding to the target servo instruction set is generated by using a first model learning module of the machine learning network model, wherein the first model learning module of the machine learning network model is used to supplement the missing instruction element data in the target servo instruction set by using optimization; The adaptive optimization instruction set is loaded into the machine learning network model, and a servo system parameter configuration result of the adaptive optimization instruction set is generated by using a second model learning module of the machine learning network model, wherein the servo system parameter information included in the servo system parameter configuration result matches the target servo working application scenario; Adaptively adjusting and optimizing the servo system parameters based on the servo system parameter configuration result of the adaptive optimization instruction set; The step of loading the target servo instruction set and the priori servo instruction set into a machine learning network model, and using a first model learning module of the machine learning network model to generate an adaptive optimization instruction set corresponding to the target servo instruction set includes: Integrate the basic network model architecture and the first special training module parameters in the machine learning network model to generate an instruction optimization sub-model, where the instruction optimization sub-model corresponds to the first model learning module of the machine learning network model; The target servo instruction set, the priori servo instruction set and the first training guidance information are loaded into the instruction optimization sub-model, and the adaptive optimization instruction set corresponding to the target servo instruction set is generated by using the instruction optimization sub-model, wherein the first training guidance information is used to control the instruction optimization sub-model to optimize the target servo instruction set; The step of loading the adaptive optimization instruction set into the machine learning network model and using the second model learning module of the machine learning network model to generate a servo system parameter configuration result of the adaptive optimization instruction set includes: Integrate the basic network model architecture and the second special training module parameters in the machine learning network model to generate a parameter configuration sub-model, where the parameter configuration sub-model corresponds to the second model learning module of the machine learning network model; The adaptive optimization instruction set and the second training guidance information are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration result of the adaptive optimization instruction set. The second training guidance information is used to control the parameter configuration sub-model to construct the servo system parameter information from the adaptive optimization instruction set.

2. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 1, characterized in that: The step of loading the adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model and using the parameter configuration sub-model to generate the servo system parameter configuration result of the adaptive optimization instruction set includes: Loading the adaptive optimization instruction set and the second training guidance information corresponding to the target servo work application scenario into the parameter configuration sub-model, and using the parameter configuration sub-model to generate the servo system parameter configuration result of the adaptive optimization instruction set, wherein different servo work application scenarios correspond to different training guidance information; Or, the adaptive optimization instruction set and the second training guidance information corresponding to multiple servo work application scenarios are loaded into the parameter configuration sub-model, and the parameter configuration sub-model is used to generate the servo system parameter configuration result of the adaptive optimization instruction set, wherein the second training guidance information is shared training guidance information of the multiple servo work application scenarios, and the target servo work application scenario belongs to the multiple servo work application scenarios.

3. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 2, characterized in that: The method further comprises: Acquire template parameter configuration learning data for different servo work application scenarios, wherein the template parameter configuration learning data includes a template adaptive optimization instruction set and annotated servo system parameter information corresponding to the template adaptive optimization instruction set; Loading the template adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model, and using the parameter configuration sub-model to generate a template servo system parameter configuration result of the template adaptive optimization instruction set; Determining a parameter configuration error according to the template servo system parameter configuration result and the marked servo system parameter information; Optimize the second special training module parameters and the second training guidance information according to the parameter configuration error.

4. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 3 is characterized in that: The step of loading the template adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model, and using the parameter configuration sub-model to generate a template servo system parameter configuration result of the template adaptive optimization instruction set includes: Loading the template adaptive optimization instruction set and the second training guidance information corresponding to the servo work application scenario to which the template adaptive optimization instruction set belongs into the parameter configuration sub-model, and using the parameter configuration sub-model to generate the template servo system parameter configuration result of the template adaptive optimization instruction set; The step of optimizing the second special training module parameters and the second training guidance information according to the parameter configuration error includes: The second special training module parameters and the second training guidance information corresponding to different servo work application scenarios are optimized according to the parameter configuration error.

5. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 4, characterized in that: The method further comprises: According to the correlation between the servo work application scenario and the instruction elements in the instruction knowledge set, x reference instruction elements are determined from the instruction knowledge set, the correlation between the reference instruction elements and the servo work application scenario is greater than the correlation between other instruction elements and the servo work application scenario, and the instruction knowledge set is obtained by iterative learning in the process of optimizing the model parameters of the basic network model architecture; Initializing the second training guidance information corresponding to the servo work application scenario according to the x reference instruction elements and the random instruction elements; The step of initializing the second training guidance information corresponding to the servo work application scenario according to the x reference instruction elements and the random instruction elements includes: Performing attribute analysis on the input x reference instruction elements, and dividing the x reference instruction elements into different attribute categories according to the attribute analysis result; Analyze the position, functional tendency and potential association of the random instruction element with other known instruction elements in the entire instruction system to determine the attribute supplement direction information of the random instruction element in supplementing the training guidance information, wherein the attribute supplement direction information is used to guide the fusion process of the random instruction element and the reference instruction element; According to the attribute categories of the reference instruction elements, a preliminary framework structure is constructed, wherein the preliminary framework structure is used to express the relationship between different attribute categories and the logical order of different attribute categories in the servo work application scenario; Supplementing direction information according to the attributes of the random instruction element, reserving a position for integrating the random instruction element in the preliminary framework structure, and generating an intermediate framework structure including the classification relationship of the reference instruction element and the position where the random instruction element is integrated; According to the importance of different parts in the intermediate framework structure and the degree of influence in the servo work application scenario, the rules for setting the weight of the instruction element are determined, wherein, for the random instruction element, the range and adjustment direction of the initial weight setting of the random instruction element are determined according to the attribute supplement direction information and the integration position in the intermediate framework structure; According to the rule for setting the weight of the instruction element, weights are assigned to the reference instruction element and the random instruction element, and an instruction element set including all the reference instruction elements and the random instruction elements and having weight information is generated; Based on the set of instruction elements with weights, an instruction element association matrix is ​​constructed, wherein the rows and columns of the instruction element association matrix are instruction elements, and the elements in the instruction element association matrix represent the degree of association between two instruction elements, and the determination of the degree of association is based on the weights, attribute categories and logical relationships of the instruction elements in the framework structure; The degree of association in the instruction element association matrix is ​​converted into the priority order and influencing factor of different instruction elements when constructing parameter information. At the same time, according to the weight information of the instruction elements, the importance of different instruction elements in the training guidance information is determined to construct and generate the second training guidance information.

6. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 3, characterized in that: The step of loading the template adaptive optimization instruction set and the second training guidance information into the parameter configuration sub-model, and using the parameter configuration sub-model to generate a template servo system parameter configuration result of the template adaptive optimization instruction set includes: Loading the template adaptive optimization instruction set and the second training guidance information corresponding to a plurality of servo work application scenarios into the parameter configuration sub-model, and using the parameter configuration sub-model to generate the template servo system parameter configuration result of the template adaptive optimization instruction set; The step of optimizing the second special training module parameters and the second training guidance information according to the parameter configuration error includes: The second special training module parameters and the second training guidance information corresponding to the multiple servo work application scenarios are optimized according to the parameter configuration error.

7. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 6, characterized in that: The method further comprises: According to the correlation between the servo work application scenario and the instruction elements in the instruction knowledge set, x reference instruction elements are determined from the instruction knowledge set, the correlation between the reference instruction elements and the servo work application scenario is greater than the correlation between other instruction elements and the servo work application scenario, and the instruction knowledge set is obtained by iterative learning in the process of optimizing the model parameters of the basic network model architecture; Fusion of the x reference instruction elements corresponding to different servo work application scenarios to generate fusion instruction elements corresponding to the multiple servo work application scenarios; The second training guidance information corresponding to the plurality of servo work application scenarios is initialized according to the fusion instruction element and the random instruction element.

8. The method for adaptive parameter adjustment and optimization of a servo system based on machine learning according to claim 1, characterized in that: The method comprises: Acquire template instruction optimization learning data for different servo work application scenarios, wherein the template instruction optimization learning data includes a template servo instruction set, a template priori servo instruction set, and a labeled adaptive optimization instruction set corresponding to the template servo instruction set; Loading the template servo instruction set, the template priori servo instruction set and the first training guidance information into the instruction optimization sub-model, and using the instruction optimization sub-model to generate a template adaptive optimization instruction set corresponding to the template servo instruction set; Determining an instruction optimization error according to the template adaptive optimization instruction set and the annotation adaptive optimization instruction set; Optimizing the first special training module parameters according to the instruction optimization error; The step of obtaining template instruction optimization learning data for different servo work application scenarios includes: Loading the template servo instruction set and the template a priori servo instruction set into a first machine learning network, and using the first machine learning network to generate a first servo system parameter configuration result of the template servo instruction set; Loading the instruction optimization sample, the template servo instruction set and the template priori servo instruction set into a second machine learning network, and using the second machine learning network to generate a reference adaptive optimization instruction set corresponding to the template servo instruction set; The template instruction optimization learning data for different servo work application scenarios is generated based on the comparison result between the first servo system parameter configuration result and the reference adaptive optimization instruction set.

9. A servo system adaptive parameter adjustment and optimization system based on machine learning, characterized in that: The servo system adaptive parameter adjustment and optimization system based on machine learning includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the servo system adaptive parameter adjustment and optimization method based on machine learning as described in any one of claims 1 to 8.

Images