Controller and control method for intelligent manufacturing of industrial tool body

Abstract

The invention discloses a controller and a control method for intelligent manufacturing of an industrial tool, and the controller comprises a sensing module which is used for obtaining multi-modal sensing data in the execution process of an industrial production task; the calculation module is used for reasoning current attribute data of at least one target object in the industrial production task according to the multi-modal sensing data, and generating a corresponding control decision according to the current attribute data of each target object and preset target attribute data corresponding to each target object at present; the control module is used for converting the control decision into a control signal for performing function control on the intelligent machine equipment; and the driving module is used for driving the intelligent machine equipment to execute corresponding operation in the industrial production task according to the control signal. According to the technical scheme in the embodiment of the invention, intelligent and efficient production of industrial manufacturing is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent controllers, and in particular to a controller and a control method for industrial embodied intelligent manufacturing. Background Art

[0002] In the field of industrial manufacturing, traditional controllers are designed based on fixed programming logic and preset rules, and it is difficult to automatically and flexibly adjust control strategies in complex process scenarios. In the coordination of multiple devices and the overall optimization of production lines, many mechanical equipment, sensors and automation devices need to work closely together, but traditional controllers can only control a single machine and it is difficult to optimize and schedule globally. Traditional controllers lack autonomous perception and decision-making capabilities in complex industrial environments, and are difficult to adapt to dynamic changes. They do not have edge computing and intelligent analysis functions, and are unable to quickly process data and analyze production situations locally. It is difficult to respond to production site needs in a timely manner, and it is unable to meet the urgent needs of intelligent and efficient production in industrial manufacturing, becoming an important bottleneck for industrial upgrading. Summary of the invention

[0003] The present invention provides a controller for industrial embodied intelligent manufacturing and a control method thereof, so as to at least solve the deficiencies existing in the above-mentioned prior art.

[0004] According to one aspect of the present invention, a controller for industrial embodied intelligent manufacturing is provided, comprising: The perception module is used to obtain multimodal sensor data during the execution of industrial production tasks; A calculation module, used to infer the current attribute data of at least one target object in the industrial production task according to the multimodal sensing data, and generate a corresponding control decision according to the current attribute data of each target object and the preset target attribute data currently corresponding to each target object; A control module, used to convert the control decision into a control signal for controlling the function of the intelligent machine device; A driving module is used to drive the intelligent machine device to perform corresponding operations in the industrial production task according to the control signal.

[0005] Further, the computing module includes: an inference unit and a decision unit; The inference unit is used to preprocess the multimodal sensor data to obtain standardized sensor data, and input the standardized sensor data into a preconfigured inference model to obtain the corresponding current attribute data of at least one target object; the inference model is a pre-trained model that can perform inference based on the input standardized sensor data and output the current attribute data of at least one target object in the industrial production task; The decision-making unit is used to generate a corresponding control decision according to the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0006] Furthermore, the inference unit is further used to load a model configuration file from an edge device, wherein the model configuration file includes the inference model and a data format of standardized sensor data input to the inference model; The data formats include: high-precision FP32 format and low-precision INT8 format.

[0007] Further, the decision unit is specifically used to generate a first control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is within a preset reasonable difference range; and generate a second control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is outside the preset reasonable difference range; Among them, one of the first control decision and the second control decision is used to control the intelligent machine device to perform the target operation, and the other of the two is used to control the intelligent machine device not to perform the target operation.

[0008] Furthermore, the decision unit is specifically used to input the current attribute data of each target object and the preset target attribute data corresponding to each target object at present into a pre-configured reinforcement learning model to obtain a corresponding control decision; The control decision can control the intelligent machine equipment to perform corresponding operations so that each target object generates corresponding changed attribute data. The goal of the reinforcement learning model is to make the current attribute data of each target object plus the corresponding changed attribute data reach the corresponding preset target attribute data as much as possible.

[0009] Furthermore, the reasoning model is selected from one of a mechanism model, a neural network model, and a mechanism-neural network model; The mechanism model is used to perform data conversion on the input standardized sensor data according to the process principle and constraint conditions to generate mechanism constraint data, which is used as the current attribute data of at least one target object; The neural network model is used to perform feature recognition on the input standardized sensor data, generate high-dimensional feature representation data, and use it as the current attribute data of at least one target object; The mechanism-neural network model is used to first screen and convert the input standardized sensor data to generate mechanism constraint data; then perform feature recognition on the mechanism constraint data to obtain high-dimensional feature representation data, and use it as the current attribute data of at least one target object.

[0010] Furthermore, the control module is specifically used to convert the optimal control decision into functional control of the intelligent machine equipment, construct a functional control program for each functional control through a programming language; and generate a control signal when the functional control program is running; The functional control includes: at least one of logic control, timing control, counting control, analog quantity control, closed-loop control, position control, speed control and sequence control; The programming language includes at least one of an instruction list, a structured text, a function block diagram, a ladder diagram, and a sequential function chart.

[0011] Further, the control signal includes: at least one of a digital control instruction, a sensor analog signal, and a servo motor configuration signal; The driving module includes: a digital signal driving unit, an analog signal driving unit and a servo control unit; The digital signal driving unit is used to convert the digital control instruction into a driving signal recognizable by the intelligent machine device; The analog signal driving unit is used to convert the sensor analog signal into a control input recognizable by the intelligent machine device; The servo control unit is used to convert the servo motor configuration signal into a servo motor configuration that can be recognized by the intelligent machine device.

[0012] Furthermore, it also includes: A communication module is used to communicate with a cloud server to receive a model configuration file generated by the cloud server and store the model configuration file at an edge device; the cloud server is used to collect execution results of the intelligent machine device and generate a new model configuration file according to the execution results; The communication module supports wired, wireless and IoT communications, and is compatible with Modbus and OPC UA protocols.

[0013] According to another aspect of the present invention, a control method for a controller for industrial embodied intelligent manufacturing is provided, comprising: S1. Acquire multimodal sensor data during the execution of industrial production tasks; S2. Inferring the current attribute data of at least one target object in the industrial production task according to the multimodal sensing data, and generating a corresponding control decision according to the current attribute data of each target object and the preset target attribute data currently corresponding to each target object; S3, converting the control decision into a control signal for controlling the function of the intelligent machine device; S4. Drive the intelligent machine device to perform corresponding operations in the industrial production task according to the control signal.

[0014] According to another aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the computer program.

[0015] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The technical solution in the embodiment of the present invention can quickly adapt to environmental changes, optimize the collaborative work between intelligent machines and equipment, and significantly improve product quality through the powerful reasoning and decision-making functions of the computing module; the cooperation between the perception module and the control module ensures fine control of the production process, improves human-machine collaboration, reduces operational difficulty, enhances flexibility, and enhances system compatibility and scalability, bringing significant economic benefits and competitiveness to enterprises. It can greatly improve industrial production efficiency.

[0017] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 is a schematic diagram of the structure of a controller for industrial embodied intelligent manufacturing provided according to an embodiment of the present invention; Figure 2 is a schematic flow chart of a cold rolling process applicable to an embodiment of the present invention; Figure 3 is a flow chart of a control method of a controller for industrial embodied intelligent manufacturing provided according to an embodiment of the present invention; Figure 4 It is a structural schematic diagram of an electronic device for implementing a control method of a controller for industrial embodied intelligent manufacturing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0021] It should be noted that it should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] Embodiment 1: Figure 1 is a schematic diagram of the structure of a controller for industrial embodied intelligent manufacturing provided according to an embodiment of the present invention, such as Figure 1 As shown, the controller includes: a sensing module 100 , a computing module 200 , a control module 300 , a driving module 400 , and a communication module 500 .

[0024] The perception module 100 is used to obtain multimodal sensor data during the execution of industrial production tasks.

[0025] In the embodiment of the present invention, industrial production tasks include: steel smelting, cold rolling production and other industrial production operations. The perception module 100 can perform sensor data fusion on data collected from different sensors to obtain multimodal sensing data. Among them, the types of sensors include: tactile sensors, temperature sensors, tilt sensors, pressure sensors, pH sensors, visual sensors, displacement sensors, speed sensors, atmosphere sensors, tension sensors, and position sensors; multimodal sensing data includes: various sensing signals such as switch signals, pulse signals, analog signals, audio signals, and image signals.

[0026] Sensor fusion is the process of integrating data from multiple sensors to provide more accurate and reliable information. By carrying out all-round, multi-modal information collection at the industrial production site, comprehensive and accurate information collection can be achieved. In robot navigation, sensor fusion can help the robot locate its own position and environment more accurately, thereby improving the efficiency and safety of navigation. In industrial automation systems, sensor fusion can monitor various parameters in the production process in real time to ensure the stability and safety of the production process.

[0027] The computing module 200 is used to infer the current attribute data of at least one target object in the industrial production task based on the multimodal sensing data, and generate corresponding control decisions based on the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0028] It should be noted that target objects refer to various reference objects that can be directly or indirectly observed in industrial production tasks; attribute data refer to reference attributes that can directly or indirectly reflect production information in industrial production tasks. For example, in steel smelting, the target object can be molten steel, etc., and the attribute data can be the temperature range and composition distribution of molten steel, etc.; in cold rolling production, the target object can be acid, strip steel, rolls, etc., and the attribute data can be the consumption rate of acid, the pickling effect of strip steel, the deformation law of strip steel, the organizational transformation process of strip steel, the type of surface defects of strip steel, the wear of rolls, the stability of rolls, etc.

[0029] Since the attribute data of most target objects cannot be directly obtained through sensors or mathematical calculations, such as the pickling effect of strip steel, the wear of rolls, the stability of rolls, etc. Therefore, the technical solution in the embodiment of the present invention cleverly designs a variety of reasoning models, which can infer the current attribute data of at least one target object in the industrial production task through reasoning based on the multimodal sensor data obtained during the execution of the industrial production task.

[0030] In the embodiment of the present invention, the computing module 200 can perform high-speed parallel computing and has reasoning and decision-making functions, including: a reasoning unit 2001 and a decision-making unit 2002 .

[0031] The reasoning unit 2001 is used to preprocess the multimodal sensor data to obtain standardized sensor data, and input the standardized sensor data into a pre-configured reasoning model to obtain the corresponding current attribute data of at least one target object; the reasoning model is a pre-trained model that can perform reasoning based on the input standardized sensor data and output the current attribute data of at least one target object in the industrial production task.

[0032] In a preferred embodiment, the inference unit 2001 is also used to load a model configuration file from the edge device, the model configuration file including the inference model and the data format of the standardized sensor data input to the inference model, wherein the data format includes: high-precision FP32 format and low-precision INT8 format.

[0033] In an embodiment of the present invention, the inference unit 2001 supports mixed precision operations, including FP32, INT8, etc. FP32 provides high-precision calculations and is suitable for key algorithms with strict precision requirements. INT8 operates with lower precision, greatly reducing the amount of calculation and storage requirements, and is suitable for processing large amounts of conventional data. Through mixed precision operations, the operation precision can be flexibly switched according to different operation scenarios, while ensuring the control accuracy, the overall operation performance is greatly improved to meet the complex needs of industrial embodied intelligent manufacturing.

[0034] In a preferred embodiment, the inference model is selected from one of a mechanism model, a neural network model, and a mechanism-neural network model. The mechanism model is used to perform data conversion on the input standardized sensor data according to the process principle and constraint conditions, generate mechanism constraint data, and use it as the current attribute data of at least one target object; the neural network model is used to perform feature recognition on the input standardized sensor data, generate high-dimensional feature representation data, and use it as the current attribute data of at least one target object; the mechanism-neural network model is used to first perform data screening and conversion on the input standardized sensor data, generate mechanism constraint data, and then perform feature recognition on the mechanism constraint data to obtain high-dimensional feature representation data, and use it as the current attribute data of at least one target object.

[0035] Among them, the reasoning basis of the mechanism model is derived from the principles of industrial production, which can accurately analyze the operating laws in the production process and provide theoretical support for the control strategy; the neural network model reasoning can process complex nonlinear data, learn through a large amount of data, and adaptively adjust the weights to achieve intelligent prediction and control of the production process; the mechanism-neural network model is suitable for scenarios involving both industrial production principles and data prediction.

[0036] In the embodiment of the present invention, the mechanism model pre-stores the industrial data and industrial knowledge involved in the industrial production process. Based on the industrial data and industrial knowledge, the process principles and constraints involved in different production scenarios are determined. According to the process principles, corresponding data operations can be performed, and according to the constraints, corresponding data screening can be performed. Industrial data comes from all aspects of the product life cycle and provides a basis for control decisions; industrial knowledge covers professional content in multiple fields. Industrial data and industrial knowledge together lay the foundation for data insights and digital representation to improve industrial process productivity, efficiency and innovation.

[0037] Furthermore, when the inference model is configured as a mechanism-neural network model, the mechanism constraint data enters the neural network model and performs anomaly detection: if an abnormality is detected, strategy re-planning is triggered, the constraints of the mechanism model are adjusted in reverse, and the standardized data is re-input into the mechanism model; if the detection is normal, high-dimensional feature representation data is obtained and used as the current attribute data of at least one target object.

[0038] The decision unit 2002 is used to generate a corresponding control decision according to the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0039] In a preferred embodiment, the decision unit 2002 is specifically used to generate a first control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is within a preset reasonable difference range; and to generate a second control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is outside the preset reasonable difference range. One of the first control decision and the second control decision is used to control the intelligent machine device to perform the target operation, and the other of the two is used to control the intelligent machine device not to perform the target operation.

[0040] The embodiment of the present invention can be applied to scenarios such as temperature measurement and sampling tasks, and the current attribute data of the target object will naturally tend to the preset target attribute data corresponding to the target object at present. In the temperature measurement and sampling task scenario, the steel block is smelted in the smelting furnace, and its temperature continues to increase and will naturally reach the preset target temperature range; when the temperature is within the preset target temperature range, the intelligent machine equipment is controlled to perform the target operation and perform molten steel sampling; when the temperature is not within the preset target temperature range, molten steel sampling is not performed.

[0041] In a preferred embodiment, the decision unit 2002 is specifically used to input the current attribute data of each target object and the preset target attribute data corresponding to each target object at present into a pre-configured reinforcement learning model to obtain a corresponding control decision. The control decision can control the intelligent machine device to perform a corresponding operation so that each target object generates corresponding changed attribute data. The goal of the reinforcement learning model is to make the current attribute data of each target object plus the corresponding changed attribute data reach the corresponding preset target attribute data as much as possible.

[0042] The embodiment of the present invention is applicable to scenarios such as steel strip pickling tasks, and the current attribute data of the target object will not tend to the preset target attribute data corresponding to the target object at present. In the steel strip pickling task scenario, the concentration and temperature of the acid solution will gradually deviate from the preset concentration and temperature as the steel strip pickling process progresses; the reinforcement learning model controls the intelligent machine equipment to perform corresponding operations, so that the concentration and temperature of the acid solution tend to the preset concentration and temperature.

[0043] In the embodiment of the present invention, the flexibility and adaptability of the decision unit 2002 directly affect the intelligence level of the controller, and the intelligent machine equipment is controlled in real time to perform corresponding operations according to changes in the environment and task requirements. Through the reinforcement learning model, the decision strategy is continuously learned and optimized in the actual production process, and other modules can be effectively coordinated and controlled to ensure decision efficiency to adapt to different production conditions and environmental changes.

[0044] The control module 300 is used to convert the control decision into a control signal for performing functional control on the intelligent machine device.

[0045] Furthermore, the control module 300 converts the control decision into functional control of the intelligent machine equipment, and constructs the functional control program of each functional control through programming language; when the functional control program runs, a control signal is generated; the functional control includes: logic control, timing control, counting control, analog control, closed-loop control, position control, speed control and sequence control, etc.; the programming language includes: instruction table, structured text, function block diagram, ladder diagram and sequential function diagram, etc. It has obvious advantages in development efficiency, algorithm implementation and resource management, and can cope with complex industrial scenarios.

[0046] Among them, logic, timing, and counting control can realize basic equipment operation rules and quantity statistics; analog quantity and closed-loop control can accurately adjust industrial parameters; position and speed control ensure the precise movement of intelligent machinery and equipment; and sequence control allows production steps to proceed in an orderly manner. Through diversified control methods, it can adapt to different industrial scenarios in an all-round way and significantly improve the degree of production automation and intelligence.

[0047] The driving module 400 is used to drive the intelligent machine equipment to perform corresponding operations in industrial production tasks according to the control signal.

[0048] It should be noted that the control signal includes: at least one of: a digital control instruction, a sensor analog signal, and a servo motor configuration signal. The drive module 400 includes: a digital signal drive unit 4001, an analog signal drive unit 4002, and a servo control unit 4003. Among them, the digital signal drive unit 4001 is used to convert the digital control instruction into a drive signal recognizable by the intelligent machine device; the analog signal drive unit 4002 is used to convert the sensor analog signal into a control input recognizable by the intelligent machine device; the servo control unit 4003 is used to convert the servo motor configuration signal into a servo motor configuration recognizable by the intelligent machine device.

[0049] In the embodiment of the present invention, the driving module 400 is the execution unit of the controller, which is responsible for receiving control signals and performing specific tasks. Among them, the digital signal driving unit 4001 has the ability to drive digital signals, and can efficiently and accurately convert the digital control instructions issued by the controller into driving signals that can be recognized by intelligent machine equipment, ensuring that the intelligent machine equipment operates stably according to the predetermined logic; the analog signal driving unit 4002 can adapt and amplify various analog signals from sensors, provide appropriate input for the operation of intelligent machine equipment, and ensure accurate control of analog quantities; the servo control unit 4003 can accurately control the speed, torque and position of the servo motor to achieve accurate driving of high-precision moving parts in intelligent machine equipment, meet the high requirements for the movement of intelligent machine equipment in complex industrial production processes, and effectively improve the accuracy and efficiency of production.

[0050] The communication module 500 is used to communicate with the cloud server to receive the model configuration file generated by the cloud server and store the model configuration file at the edge device; the cloud server is used to collect the execution results of the intelligent machine equipment and generate a new model configuration file based on the execution results; the communication module supports wired, wireless and Internet of Things communications, and is compatible with Modbus and OPC UA protocols.

[0051] In an embodiment of the present invention, the communication module 500 supports the edge-cloud collaboration mechanism. The cloud server collects key data from the execution results of the intelligent machine device and updates the model configuration file; the edge device deploys the updated reasoning model. By adopting the edge-cloud collaboration mechanism, the controller can make real-time adjustments when the environment changes and generate a new model configuration file. After the updated model configuration file is segmented, it is dynamically deployed to the controller and associated edge devices to form a "cloud training-edge reasoning" closed loop, continuously improving control accuracy and scene adaptability.

[0052] Preferably, transfer learning technology can also be used to quickly apply the experience and knowledge learned in one production scenario to similar production scenarios, reducing the time and cost of re-learning.

[0053] In the embodiment of the present invention, the communication module 500 also supports real-time data transmission and information interaction with external systems, ensuring that data can be quickly and accurately transmitted between different devices and systems to achieve collaborative control of the production process. The communication module 500 supports multiple communication protocols, such as Modbus, OPC UA, etc., which can achieve seamless connection between the controller and devices from different manufacturers, and improve the compatibility and scalability of the controller system.

[0054] Furthermore, the communication module 500 also supports distributed interoperability, and different communication nodes can automatically coordinate their work according to preset rules. Even if some communication nodes fail, other communication nodes can still maintain the basic operation of industrial production, greatly enhancing the reliability and fault tolerance of the industrial production system.

[0055] The technical solution of the embodiment of the present invention significantly improves product quality, greatly improves production efficiency, improves human-machine collaboration, reduces operational difficulty, and enhances flexibility through the collaborative cooperation of various modules within the controller. The powerful functions of the computing module can quickly adapt to environmental changes, optimize equipment parameters and collaborative work; the cooperation of the sensing module, control module and drive module ensures fine control of the production process; the communication module facilitates communication between external devices or systems and the controller system, enhances the compatibility and scalability of the controller, and effectively promotes industrial manufacturing towards intelligence and efficiency, bringing significant economic benefits and competitiveness to enterprises.

[0056] Embodiment 2: Embodiment 2 of the present invention is applicable to the scenario where the current attribute data of the target object will naturally tend toward the preset target attribute data corresponding to the target object at present. In combination with the temperature measurement and sampling process in the steel smelting process, Embodiment 2 of the present invention will further introduce the controller in the embodiment of the present invention. The industrial production task is set to measure the temperature and sample the molten steel, and the intelligent machine equipment is selected as a temperature measurement and sampling robot. The controller in the embodiment of the present invention is used in combination with the temperature measurement and sampling robot, so that the temperature measurement and sampling robot is endowed with powerful embodied intelligent characteristics, thereby significantly optimizing the workflow.

[0057] Perception module 100: used to obtain multimodal sensor data during the execution of industrial production tasks.

[0058] In the temperature measurement and sampling task, the multimodal sensing data includes: multispectral images collected by the visual sensor, the three-dimensional posture angle of the ladle collected by the inclination sensor, and the floating-point temperature value collected by the temperature sensor.

[0059] The various data collected by the sensors help the temperature measurement and sampling robot to accurately locate and perceive the environment. Among them, the visual sensor can accurately identify the target positions of the ladle, ingot, etc. in the high temperature and dusty steel production environment, ensuring the accurate arrival of the temperature measurement and sampling probe; the inclination sensor obtains the three-dimensional posture angle of the ladle in real time; the temperature sensor monitors the ambient temperature and the temperature of the equipment itself in real time to ensure the stable operation of the temperature measurement and sampling robot.

[0060] The reasoning unit 2001 is used to preprocess the multimodal sensor data to obtain standardized sensor data, and input the standardized sensor data into a pre-configured reasoning model to obtain the corresponding current attribute data of at least one target object; the reasoning model is a pre-trained model that can perform reasoning based on the input standardized sensor data and output the current attribute data of at least one target object in the industrial production task.

[0061] In an embodiment of the present invention, a mechanism-neural network model is selected as the inference model, which is used to first perform high-precision FP32 operations on the input floating-point temperature value and the three-dimensional posture angle of the ladle, and perform data conversion based on the thermal conductivity coefficient corresponding to the steel type in the process principle and the target temperature range of the molten steel in the constraint conditions to obtain the spatial coordinates of the safe temperature measurement area and the temperature field prediction matrix; then, low-precision INT8 operations are performed on the spatial coordinates of the safe temperature measurement area, the temperature field prediction matrix and the multi-spectral image, and the temperature of the current optimal temperature measurement point of the molten steel is obtained by identifying the dynamic characteristics of the molten pool surface.

[0062] In the embodiment of the present invention, the mechanism model is partially based on the molten steel heat conduction physical model, combined with the temperature field distribution boundary conditions, to provide reliable safety temperature measurement area constraints. The floating point temperature value of the temperature sensor and the three-dimensional attitude angle of the ladle of the inclination sensor are input into the mechanism model, and the spatial coordinates of the safety temperature measurement area and the temperature field prediction matrix (temperature distribution map of 128×128 grids) are output according to the heat conductivity coefficient corresponding to the steel grade code (Q235B) and the target temperature range of the molten steel (1550-1580℃).

[0063] The ResNet-18 model can be used as the neural network model part. The multispectral image, temperature field prediction matrix and spatial coordinates of the safe temperature measurement area can be input into the model to identify the dynamic characteristics of the molten pool surface, such as slag layer thickness, bubble position, etc., to achieve accurate point selection and output the temperature and confidence score of the current optimal temperature measurement point of the molten steel.

[0064] In addition, the spatial coordinates of the safe temperature measurement area and the temperature field prediction matrix enter the neural network model for abnormal detection: if the spatial coordinates of the safe temperature measurement area and the temperature field prediction matrix do not meet the safety range pre-set by the neural network model, the constraint conditions of the mechanism model (target temperature range of molten steel) are adjusted; if they meet the requirements, the multispectral image, the spatial coordinates of the safe temperature measurement area and the temperature field prediction matrix are used for feature recognition to obtain the temperature and coordinates of the optimal temperature measurement point. The spatial coordinates of the safe temperature measurement area and the temperature field prediction matrix output by the mechanism model limit the search space of ResNet-18 to prevent the probe from contacting the furnace wall or slag layer.

[0065] In the embodiment of the present invention, the inference unit 2001 can quickly infer the multimodal sensor data transmitted by the perception module 100 based on the mechanism-neural network model and the model trained with a large amount of steel production data. For example, the quality recognition model infers the temperature range, component distribution and other information of the molten steel based on the color, flow state and temperature change trend of the molten steel, and predicts possible quality problems in advance.

[0066] The decision unit 2002 is used to generate a corresponding control decision according to the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0067] The decision unit 2002 is specifically used to generate a first control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is within a preset reasonable difference range; and to generate a second control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is outside the preset reasonable difference range; wherein one of the first control decision and the second control decision is used to control the intelligent machine device to perform the target operation, and the other of the two is used to control the intelligent machine device not to perform the target operation.

[0068] In the embodiment of the present invention, the decision unit 2002 generates a control decision based on the difference between the temperature of the current optimal temperature measurement point of the molten steel and the preset temperature measurement sampling temperature of the molten steel, and controls the temperature measurement sampling robot to perform or not perform the target operation. For example, according to the smelting stage and process requirements of the molten steel, the appropriate temperature measurement time point, sampling depth and position are used as the target operation.

[0069] Control module 300: used to convert the control decision into a control signal for performing functional control on the intelligent machine equipment.

[0070] In the embodiment of the present invention, the control module 300 uses programming languages ​​such as ladder diagrams and statement tables to perform complex logical operations and sequential control on the actions of the temperature measurement and sampling robot to ensure precise control and stable operation. The control module 300 supports the realization of functions such as logic, timing, counting, analog quantity, closed loop, position, speed, and sequence. By setting a high-precision timer, the time interval of temperature measurement and sampling is accurately controlled to ensure the continuity and accuracy of the data; by accurately counting the pulse signal, the specific actions of the temperature measurement and sampling robot are triggered, such as: the extension and retraction of the probe, etc.; through the analog control capability, the temperature, pressure and other signals are collected and processed, and the decision instructions of the calculation module 200 are quickly converted into the logic control program instructions of the temperature measurement and sampling robot, so as to achieve precise control of the temperature measurement and sampling robot, and at the same time, an efficient closed-loop control system is constructed to feedback the operation status of the equipment in real time to ensure the stable operation of the temperature measurement and sampling robot.

[0071] Driving module 400: used to drive the intelligent machine equipment to perform corresponding operations in industrial production tasks according to the control signal.

[0072] In the embodiment of the present invention, the driving module 400 serves as an execution unit to ensure that the robot can stably and accurately complete the temperature measurement and sampling actions in a high temperature and strong magnetic field steel production environment, thereby improving the accuracy and efficiency of the work.

[0073] Among them, the digital signal drive unit 4001 efficiently converts the digital control instructions issued by the controller into drive signals that can be recognized by the temperature measurement and sampling robot; the analog signal drive unit 4002 adapts and amplifies the analog signal to provide support for precise control; the servo control unit 4003 achieves precise driving of the joints and other high-precision moving parts of the temperature measurement and sampling robot by precisely controlling the speed, torque and position of the servo motor.

[0074] Communication module 500: used to communicate with the cloud server to receive the model configuration file generated by the cloud server and store the model configuration file at the edge device; the cloud server is used to collect the execution results of the intelligent machine device and generate a new model configuration file based on the execution results.

[0075] In an embodiment of the present invention, the communication module 500 adopts industrial Ethernet or 5G technology to realize real-time data transmission and information interaction between the temperature measurement and sampling robot and other equipment in the steelmaking workshop and the upper-level monitoring system; the temperature measurement and sampling data are quickly transmitted to the quality control system so as to adjust the production process in time; at the same time, the instructions and production plan change information of the upper-level monitoring system are received to flexibly adjust the work tasks.

[0076] The technical solution in the embodiment of the present invention, through the coordinated work of the above modules, the temperature measurement and sampling robot, empowered by the controller, has embodied intelligent characteristics, can independently, efficiently and accurately complete the temperature measurement and sampling work in a complex steel production environment, and provide strong guarantee for improving the quality of steel production.

[0077] Embodiment 3: Embodiment 3 of the present invention is applicable to the case where the current attribute data of the target object does not tend to the preset target attribute data currently corresponding to the target object. In combination with the cold rolling production process, Embodiment 3 of the present invention will further introduce the controller in the embodiment of the present invention.

[0078] Figure 2 is a schematic diagram of a cold rolling process applicable to an embodiment of the present invention, such as Figure 2 As shown, the third embodiment of the present invention divides the cold rolling production into four process stages.

[0079] Perception module 100: used to obtain multimodal sensor data during the execution of industrial production tasks.

[0080] The reasoning unit 2001 is used to preprocess the multimodal sensor data to obtain standardized sensor data, and input the standardized sensor data into a pre-configured reasoning model to obtain the corresponding current attribute data of at least one target object; the reasoning model is a pre-trained model that can perform reasoning based on the input standardized sensor data and output the current attribute data of at least one target object in the industrial production task.

[0081] The first process stage: pickling process.

[0082] Sensing module 100: In the pickling process, different types of sensors are deployed in and around the pickling tank. For example, pH sensors are used to monitor the acid concentration in real time, and temperature sensors are used to control the acid temperature; high-definition visual sensors are used to check the residual oxide scale on the surface of the strip and accurately perceive the real-time status of the pickling process.

[0083] Reasoning unit 2001: A high-precision calculation mechanism model is selected, which has a small amount of calculation and low latency. The input of the mechanism model includes: strip operation parameters, acid state parameters, etc. Among them, the strip operation parameters include: real-time strip speed, strip thickness, material (Q235B), surface oxide thickness, strip temperature distribution (obtained by infrared sensor), etc.; acid state parameters include: acid concentration (pH value, Fe²⁺ ion concentration), acid temperature, etc. The process principle and constraints include: equipment and process indicators, such as: looping quantity, speed limit of each section of the unit, etc. The model output includes: acid consumption rate, strip pickling effect, etc., as the current attribute data of the target object in the first process stage.

[0084] The second process stage: rolling process.

[0085] Perception module 100: During the rolling process, pressure sensors, displacement sensors, speed sensors and high-precision plate shape detectors are installed at various key locations of the rolling mill to monitor the rolling force, position and speed of the rolls, and changes in the shape and thickness of the strip in real time, so as to fully grasp the real-time status of the rolling process.

[0086] Reasoning unit 2001: Selection and mechanism-neural network model, the mechanism model part adopts high precision, and the neural network model part adopts low precision.

[0087] In the mechanism model, the Stone rolling force model can be used. The model input includes: real-time sensor data, such as rolling force, roll gap, and strip entry / exit thickness, etc.; the setting of process principles and constraints includes: strip yield strength, friction coefficient; model output includes: theoretical rolling force range, roll gap adjustment range.

[0088] When the output data of the mechanism model is used for the neural network model, if the actual plate quality deviation is greater than 5% or the rolling force exceeds the theoretical range, the friction coefficient in the mechanism model is updated in reverse, and the ResNet-50 plate shape detection model fine-tunes the network weights at the same time.

[0089] In the neural network model, a combination of the ResNet-50 flatness detection model and the reinforcement learning model can be used. In the ResNet-50 flatness detection model, the model input includes: flatness detector image, strip thickness distribution matrix, etc.; the model output includes: flatness defect classification code and defect location thermal map, etc. In the reinforcement learning model, the model input includes: current rolling parameters, flatness quality score, equipment status (roller temperature, vibration amplitude), planned output, etc. The model output includes: the deformation law of the strip, the wear of the rolls, and the stability of the rolls during rolling, as the current attribute data of the target object in the second process stage.

[0090] The third process stage: annealing process.

[0091] Sensing module 100: In the annealing process, temperature sensors and atmosphere sensors are installed in the annealing furnace, and strip tension sensors and position sensors are installed outside the furnace. The temperature distribution and gas composition in the furnace, as well as the tension and position changes of the strip during the annealing process are monitored in real time to fully grasp the information in the annealing furnace.

[0092] Reasoning unit 2001: neural network model is selected. The input of the neural network model includes: current temperature curve (heating rate, holding time), strip performance indicators (yield strength, elongation), energy consumption data (gas consumption, electricity consumption), delivery cycle constraints (total annealing time ≤ 120 minutes), quality requirements (hardness HRB ≤ 75), etc.; the output of the model is: heat exchange in the annealing furnace, the organizational transformation process of the strip and possible defects, as the current attribute data of the target object in the third process stage.

[0093] The fourth process stage: finishing process.

[0094] Perception module 100: In the finishing process, surface defect detectors, dimension measuring instruments and flatness detectors are installed on the finishing equipment to detect the surface quality, dimension accuracy and flatness of the strip in real time, and fully understand the quality status of the strip during the finishing process.

[0095] Reasoning unit 2001: A neural network model with low-precision calculation is selected. The neural network model can be a combination of a YOLOv8 surface defect detection model and a Vision Transformer quality assessment model.

[0096] In the YOLOv8 surface defect detection model, the model input includes: spectral surface image, strip motion parameters (speed, acceleration), etc.; the model output includes: defect type and coordinates, etc.

[0097] In the Vision Transformer quality assessment model, the model inputs include: surface defect heat map, thickness / width time series curve, and quality standards; the model outputs include: the type of surface defects of the strip, the cause of their occurrence, and the trend of dimensional deviation, which serve as the current attribute data of the target object in the fourth process stage.

[0098] The decision unit 2002 is used to generate a corresponding control decision according to the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0099] The decision unit 2002 is specifically used to input the current attribute data of each target object and the preset target attribute data corresponding to each target object at present into a pre-configured reinforcement learning model to obtain a corresponding control decision; the control decision can control the intelligent machine equipment to perform corresponding operations so that each target object generates corresponding changed attribute data. The goal of the reinforcement learning model is to make the current attribute data of each target object plus the corresponding changed attribute data to reach the corresponding preset target attribute data as much as possible.

[0100] In an embodiment of the present invention, the decision unit 2002 performs global production rhythm scheduling through a reinforcement learning model, including: precise matching of pickling feed and rolling consumption rate, adjustment of acid pH, rolling force, annealing temperature and finishing shear amount to eliminate quality deviation, etc., to build a process-equipment-quality association knowledge graph. In addition, the reinforcement learning model can also reversely correct the parameters of the reasoning model of each process stage. For example: when the finishing quality score does not meet the standard, the finishing process → annealing process → rolling process → pickling process is reversely tested. When the yield strength of the steel block is detected to be insufficient, the constraints of each process stage are adjusted, for example: increasing the annealing holding time, reducing the rolling speed to improve deformation uniformity, and adjusting the pickling temperature.

[0101] Specifically, the decision unit 2002 can generate control decisions: determine whether the first process stage needs to add acid, adjust the acid temperature, or change the traveling speed of the strip; optimize the acid addition and temperature control during long-term operation to meet the pickling requirements of different batches of strip. Develop the optimal rolling force, roll speed, and roll gap adjustment plan for the second process stage; when facing the rolling tasks of strips of different specifications and materials, improve the rolling quality and efficiency, continuously improve the rolling process model, and achieve precise control of the rolling process. Develop the best heating curve, holding time, and furnace atmosphere control plan for the third process stage; improve the annealing quality and performance of the strip through continuous analysis of annealing data. Develop the best finishing process plan for the fourth process stage, such as shear length, straightening force, etc. When facing strips of different specifications and quality requirements, improve the finishing quality and production efficiency, continuously improve the finishing process model, and achieve precise control of the strip quality.

[0102] Control module 300: used to convert the control decision into a control signal for performing functional control on the intelligent machine equipment.

[0103] In the embodiment of the present invention, the control module 300 accurately controls the opening and closing of the acid addition pump, the power adjustment of the heating device, and the speed of the strip conveying motor for the first process stage. By setting a high-precision timer, the time accuracy of each link operation is ensured. By using the analog control capability, the analog signals such as acid concentration and temperature are collected and processed in real time to achieve closed-loop control of the pickling process. For the second process stage: the various actuators of the rolling mill are accurately controlled using programming languages ​​such as ladder diagrams and statement tables. The precise adjustment of the roller position is achieved by accurately counting the pulse signal. By using the analog control capability, the analog signals such as rolling force and speed are collected and processed in real time to build an efficient closed-loop control system to ensure the stable operation of the rolling mill and the rolling quality of the strip meets the standards. For the third process stage: the heating elements, gas flow control valves, strip tension adjustment devices, etc. of the annealing furnace are accurately controlled. Complex logical operations and sequential control are used to ensure that each device operates according to the predetermined annealing process curve. By setting a high-precision timer, the heating and insulation time are accurately controlled. By using analog control capabilities, we can collect and process analog signals such as temperature and gas flow in real time to achieve closed-loop control of the annealing process. For the fourth process stage: we can precisely control the actuators of the finishing equipment, such as shears and straighteners. We can use programming languages ​​such as ladder diagrams and statement tables to achieve logical and sequential control of the equipment. We can ensure the accuracy of the shearing length by accurately counting the pulse signals. By using analog control capabilities, we can collect and process analog signals such as straightening force in real time to build an efficient closed-loop control system to ensure the finishing quality of the strip.

[0104] Driving module 400: used to drive the intelligent machine equipment to perform corresponding operations in industrial production tasks according to the control signal.

[0105] The technical solution in the embodiment of the present invention, through the distributed application controllers in each process section of the cold rolling production line, each process section has the ability of autonomous perception, intelligent decision-making and precise control. The intelligent controllers of each process section exchange data and work together in real time through the communication module, and combined with the unified scheduling of the upper monitoring system, the entire cold rolling production line forms an organic intelligent whole, realizing a highly intelligent transformation, and greatly improving production efficiency, product quality and enterprise competitiveness.

[0106] Embodiment 4: Figure 3 is a flow chart of a control method of a controller for industrial embodied intelligent manufacturing provided according to an embodiment of the present invention, such as Figure 3 As shown, the control method includes: S1. Acquire multimodal sensor data during the execution of industrial production tasks.

[0107] S2. Infer the current attribute data of at least one target object in the industrial production task based on the multimodal sensing data, and generate corresponding control decisions based on the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

[0108] S3. Convert the control decision into a control signal for controlling the function of the intelligent machine equipment.

[0109] S4. Drive the intelligent machine equipment to perform corresponding operations in industrial production tasks according to the control signal.

[0110] Preferably, it also includes: S5. Communicate with the cloud server to receive the model configuration file generated by the cloud server, and store the model configuration file at the edge device; the cloud server is used to collect the execution results of the intelligent machine device and generate a new model configuration file based on the execution results.

[0111] The technical solution in the embodiment of the present invention proposes a control method for a controller for industrial embodied intelligent manufacturing. Based on a distributed edge intelligent architecture, it deeply integrates a closed-loop collaborative mechanism of perception, computing, control, and execution to achieve dynamic optimization and real-time response of industrial embodied intelligent manufacturing.

[0112] Embodiment five: Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0113] like Figure 4As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0114] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0115] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a control method for a controller for industrial embodied intelligent manufacturing.

[0116] In some embodiments, the control method of the controller for industrial embodied intelligent manufacturing can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the control method of the controller for industrial embodied intelligent manufacturing described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the control method of the controller for industrial embodied intelligent manufacturing in any other appropriate manner (for example, by means of firmware).

[0117] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0118] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0119] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0120] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0121] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0122] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0123] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0124] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A controller for industrial embodied intelligent manufacturing, characterized in that: include: The perception module is used to obtain multimodal sensor data during the execution of industrial production tasks; A calculation module, used to infer the current attribute data of at least one target object in the industrial production task according to the multimodal sensing data, and generate a corresponding control decision according to the current attribute data of each target object and the preset target attribute data currently corresponding to each target object; A control module, used to convert the control decision into a control signal for controlling the function of the intelligent machine device; A driving module is used to drive the intelligent machine device to perform corresponding operations in the industrial production task according to the control signal.

2. The controller according to claim 1, characterized in that: The computing module includes: an inference unit and a decision unit; The inference unit is used to preprocess the multimodal sensor data to obtain standardized sensor data, and input the standardized sensor data into a preconfigured inference model to obtain the corresponding current attribute data of at least one target object; the inference model is a pre-trained model that can perform inference based on the input standardized sensor data and output the current attribute data of at least one target object in the industrial production task; The decision-making unit is used to generate a corresponding control decision according to the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object.

3. The controller according to claim 2, characterized in that: The inference unit is further used to load a model configuration file from an edge device, wherein the model configuration file includes the inference model and a data format of standardized sensor data input to the inference model; The data formats include: high-precision FP32 format and low-precision INT8 format.

4. The controller according to claim 2, characterized in that: The decision unit is specifically configured to generate a first control decision when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is within a preset reasonable difference range; And, when the difference between the current attribute data of each target object and the preset target attribute data currently corresponding to each target object is outside the preset reasonable difference range, a second control decision is generated; Among them, one of the first control decision and the second control decision is used to control the intelligent machine device to perform the target operation, and the other of the two is used to control the intelligent machine device not to perform the target operation.

5. The controller according to claim 2, characterized in that: The decision-making unit is specifically used to input the current attribute data of each target object and the preset target attribute data corresponding to each target object at present into a pre-configured reinforcement learning model to obtain a corresponding control decision; The control decision can control the intelligent machine equipment to perform corresponding operations so that each target object generates corresponding changed attribute data. The goal of the reinforcement learning model is to make the current attribute data of each target object plus the corresponding changed attribute data reach the corresponding preset target attribute data as much as possible.

6. The controller according to claim 2, characterized in that: The reasoning model is selected from one of a mechanism model, a neural network model, and a mechanism-neural network model; The mechanism model is used to perform data conversion on the input standardized sensor data according to the process principle and constraint conditions to generate mechanism constraint data, which is used as the current attribute data of at least one target object; The neural network model is used to perform feature recognition on the input standardized sensor data, generate high-dimensional feature representation data, and use it as the current attribute data of at least one target object; The mechanism-neural network model is used to first screen and convert the input standardized sensor data to generate mechanism constraint data; then perform feature recognition on the mechanism constraint data to obtain high-dimensional feature representation data, and use it as the current attribute data of at least one target object.

7. The controller according to claim 1, characterized in that: The control module is specifically used to convert the optimal control decision into functional control of the intelligent machine equipment, construct a functional control program for each functional control through a programming language; and generate a control signal when the functional control program is running; The functional control includes: at least one of logic control, timing control, counting control, analog quantity control, closed-loop control, position control, speed control and sequence control; The programming language includes at least one of an instruction list, a structured text, a function block diagram, a ladder diagram, and a sequential function chart.

8. The controller according to claim 1 or 6, characterized in that: The control signal includes: at least one of a digital control instruction, a sensor analog signal, and a servo motor configuration signal; The driving module includes: a digital signal driving unit, an analog signal driving unit and a servo control unit; The digital signal driving unit is used to convert the digital control instruction into a driving signal recognizable by the intelligent machine device; The analog signal driving unit is used to convert the sensor analog signal into a control input recognizable by the intelligent machine device; The servo control unit is used to convert the servo motor configuration signal into a servo motor configuration that can be recognized by the intelligent machine device.

9. The controller according to claim 1, characterized in that: Also includes: A communication module, used to communicate with the cloud server to receive a model configuration file generated by the cloud server, and store the model configuration file at the edge device; The cloud server is used to collect the execution results of the intelligent machine device and generate a new model configuration file according to the execution results; The communication module supports wired, wireless and IoT communications, and is compatible with Modbus and OPC UA protocols.

10. A control method for a controller for industrial embodied intelligent manufacturing, applied to the controller for industrial embodied intelligent manufacturing as claimed in any one of claims 1 to 9, characterized in that: The method comprises: S1. Acquire multimodal sensor data during the execution of industrial production tasks; S2. Inferring the current attribute data of at least one target object in the industrial production task according to the multimodal sensing data, and generating a corresponding control decision according to the current attribute data of each target object and the preset target attribute data currently corresponding to each target object; S3, converting the control decision into a control signal for controlling the function of the intelligent machine device; S4. Drive the intelligent machine device to perform corresponding operations in the industrial production task according to the control signal.

Images