Industrial equipment voice control method and system based on large language model

Abstract

The application discloses an industrial equipment voice control method and system based on a large language model, and belongs to the technical field of industrial automation control and artificial intelligence interaction. The method first analyzes voice control requirements of the industrial equipment, and constructs a standardized operation instruction set; acquires process knowledge elements, and constructs a structured process knowledge set; then, based on the standardized operation instruction set and the structured process knowledge set, the large language model is used to convert a voice signal input by a user into a standardized operation instruction and drive the industrial equipment to perform an action. In this way, problems such as non-standard user voice expression, many professional terms and complex control logic in industrial scene voice control are solved, important support is provided for improving operation safety, improving production efficiency and promoting upgrading of industrial production to unmanned and intelligent.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation control and artificial intelligence interaction technology, specifically relating to a voice control method and system for industrial equipment based on a large language model. Background Technology

[0002] Voice control for industrial equipment is an important new human-machine interaction method in the field of intelligent manufacturing. It eliminates the physical limitations imposed on operators by traditional physical control methods, allowing them to perform core operations such as equipment operation, command issuance, and status inquiries through natural language, even when their hands are occupied, their line of sight is limited, or in high-intensity work scenarios. This technology features high precision and low latency, significantly improving the accuracy of work execution and the level of safety in industrial production, and is a key technology driving the upgrade of industrial production towards unmanned and intelligent operations. Unlike traditional voice control, industrial voice control needs to be designed for the complex and harsh working environments of industrial sites, placing more stringent requirements on recognition accuracy, response latency, and operational stability. Traditional methods mainly fall into two categories: template matching-based methods and neural network-based methods. Template matching-based methods rely too heavily on pre-set structures and are difficult to adapt to the diverse expressions of natural language; neural network-based methods require high-quality labeled data, and their generalization ability and recognition accuracy are often constrained by the quality of industrial data. Large language models, with their strong general language processing capabilities, offer a new solution for voice control of industrial equipment. However, existing methods based on large language models lack the integration and application of industrial process knowledge, making it difficult to understand users' voice commands from a professional perspective, thus causing deviations in operation instructions. Furthermore, existing methods mainly target voice control scenarios with relatively simple commands; voice control of industrial equipment with complex logic remains unresolved.

[0003] Chinese patent CN118737141B, entitled "A Human-Computer Voice Interaction Control Method and System Based on a Smart TV," describes a method for controlling the TV by extracting user-input voice and comparing it with a preset command library. However, this method requires pre-constructing a complete set of rule templates, but it doesn't provide a method for doing so. Furthermore, the template-based matching method lacks semantic understanding, relying solely on field-based format matching, necessitating the creation of numerous templates to cover a wide range of scenarios.

[0004] Chinese patent CN119049460B, entitled "A Real-Time Voice Interaction Method and System for Robots Based on Spiking Neural Networks," describes a method for controlling a robot via voice by inputting user voice into a trained neural network for inference and outputting operational commands. However, the performance of this method heavily relies on the quality of the labeled dataset, and it does not provide a method for constructing such a dataset. For industrial equipment, the sample size is often small, making dataset construction difficult.

[0005] Chinese patent CN121545519A, entitled "Voice Control Method, Device, Electronic Equipment, Computer-Readable Storage Medium, and Computer Program Product," describes a method where collected audio is uploaded to a cloud server, a large cloud-based model is used for speech recognition, and the model interprets user intent based on pre-set prompts to generate operation commands, which are then sent to the terminal device for execution. However, this method lacks model guidance, making it difficult for the large model to understand complex conditional commands. Furthermore, it fails to incorporate process knowledge, making it difficult to understand user speech from a professional perspective, potentially leading to model illusions and deviations in operation commands. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to propose a voice control method and system for industrial equipment based on a large language model. This method addresses challenges such as non-standard user voice expressions, numerous technical terms, and complex control logic in industrial voice control scenarios, providing crucial support for improving operational safety, increasing production efficiency, and promoting the upgrading of industrial production towards unmanned and intelligent operation.

[0007] The technical solution of this invention is as follows:

[0008] On the one hand, this invention provides a voice control method for industrial equipment based on a large language model, comprising the following steps:

[0009] Analyze the voice control requirements of industrial equipment and construct a standardized set of operating instructions;

[0010] Acquire process knowledge elements and construct a structured process knowledge set; the process knowledge elements are used to describe process knowledge related to industrial equipment;

[0011] Based on a standardized set of operating instructions and a structured set of process knowledge, a large language model is used to convert user-input voice signals into standardized operating instructions and drive industrial equipment to perform actions.

[0012] Furthermore, the analysis of the voice control requirements of industrial equipment and the construction of a standardized operation instruction set specifically include:

[0013] A1: Extract keywords for voice control of industrial equipment and establish a keyword library for operation instructions;

[0014] Keywords are categorized into three types: controlled object keywords, execution action keywords, and conditional keywords. Corresponding thesauruses are then constructed to collectively form the operation instruction keyword library.

[0015] (1) Controlled object thesaurus: This includes several controlled object keywords, representing the controlled objects in industrial equipment;

[0016] (2) Execution action vocabulary: This includes several execution action keywords, which are used to describe the specific control actions performed on the controlled objects in industrial equipment;

[0017] (3) Conditional terminology: includes several conditional keywords that describe the conditions for performing an action;

[0018] A2: Legally combine keywords from the operation instruction keyword library to construct an operation instruction set; the operation instruction set includes several operation instructions, which include two types: basic operation instructions and conditional operation instructions.

[0019] (1) The basic operation instructions are formed by combining the execution action keywords and the controlled object keywords. The structure is execution action keywords + controlled object keywords;

[0020] (2) Conditional operation instructions are based on basic operation instructions, with the introduction of a conditional word library to extend the execution method of basic operation instructions. The structure of conditional operation instructions is conditional keyword + execution action keyword + controlled object keyword;

[0021] A3: Define each operation instruction in the operation instruction set in a structured form to obtain structured operation instructions, and then obtain a standardized operation instruction set;

[0022] The structured operation instructions include a controlled object field, an execution action field, and a condition field. The values ​​of the controlled object field, execution action field, and condition field are derived from corresponding dictionaries and constructed according to the operation instructions in the operation instruction set. If no condition keyword is included, the condition field is empty.

[0023] Furthermore, the acquisition of process knowledge elements and the construction of a structured process knowledge set specifically involves:

[0024] First, process knowledge elements are extracted and semantically decomposed into multiple process knowledge units. Each process knowledge unit includes a knowledge identifier field, a knowledge content field, a knowledge type field, and a source field. The knowledge identifier field is used to identify different process knowledge units, the knowledge content field is used to store specific process knowledge content, the knowledge type field is used to identify the category to which the process knowledge content belongs, and the source field is used to identify the type of document or record from which the process knowledge content originates. Then, the process knowledge units are organized in a structured manner to form a structured process knowledge set.

[0025] Furthermore, the process of converting user-input voice signals into standardized operating instructions and driving industrial equipment to perform actions based on a standardized set of operating instructions and a structured set of process knowledge, using a large language model, includes the following steps:

[0026] B1: Acquire the user's original voice signal, perform speech recognition, and convert it into text information;

[0027] B2: Retrieve process knowledge units associated with text information from the structured process knowledge set to form process knowledge information;

[0028] B3: Based on a preset thought chain template, the large language model is guided to reason based on textual information and process knowledge, output operation instructions, and drive industrial equipment to perform actions.

[0029] Furthermore, B3 specifically includes the following steps:

[0030] B3.1: Construct model prompt words based on textual information and process knowledge information;

[0031] First, textual information is concatenated with process knowledge information to form an enhanced semantic description; then, the enhanced semantic description is structured and organized to obtain model prompt words.

[0032] B3.2: Perform multi-intent recognition on model prompts and divide the model prompts into several independent statements according to the recognized control intent;

[0033] Specifically: determine whether the original speech signal contains multiple independent control intentions; when the original speech signal contains multiple control intentions, split the model prompt words into several independent sentences according to the sentence order;

[0034] B3.3: For each independent statement, identify the keywords of the controlled object, and take the industrial equipment pointed to by the identified keywords of the controlled object as the control target;

[0035] B3.4: For each individual statement, identify the keywords for the execution action;

[0036] B3.5: For each individual statement, identify conditional keywords;

[0037] B3.6: Integrate all identified keywords and determine the validity of the control intent. If neither the controlled object keyword nor the execution action keyword is identified, the control intent is deemed invalid. If at least the controlled object keyword and the execution action keyword are identified, the control intent is considered valid.

[0038] B3.7: After confirming the validity of the control intent, the extracted keywords are matched with the standardized operating instruction set to obtain the standardized operating instructions;

[0039] The specific processing method is as follows:

[0040] (1) Determining the type of operation instruction: Based on the extracted keywords, determine the type of operation instruction to be matched: when the keywords only contain keywords of the controlled object and keywords of the action to be executed, determine the basic operation instruction to be matched; when the keywords further contain conditional keywords, determine the conditional operation instruction to be matched.

[0041] (2) Operation instruction field mapping: Controlled object keywords are mapped to controlled object fields, execution action keywords are mapped to execution action fields, and condition keywords are mapped to condition fields, forming standardized operation instruction matching items;

[0042] (3) Operation instruction search and determination: In the standardized operation instruction set, search for operation instructions that have the same field values ​​as the standardized operation instruction matching items. When there is a unique matching operation instruction, determine that operation instruction as the target operation instruction; when no matching operation instruction is found, the output command is invalid.

[0043] B3.8: Outputs standardized operating instructions to drive industrial equipment to perform actions.

[0044] On the other hand, the present invention also provides an industrial equipment voice control system based on a large language model, for implementing an industrial equipment voice control method based on a large language model, comprising:

[0045] The instruction set construction module is used to analyze the voice control requirements of industrial equipment and build a standardized operation instruction set.

[0046] The process knowledge module is used to acquire process knowledge elements and construct a structured process knowledge set.

[0047] The retrieval and control module is used to convert user-input voice signals into standardized operating instructions based on a standardized set of operating instructions and a structured set of process knowledge, and to drive industrial equipment to perform actions using a large language model.

[0048] Thirdly, this application proposes an electronic device, including: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to perform the aforementioned industrial equipment voice control method based on a large language model.

[0049] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned industrial equipment voice control method based on a large language model.

[0050] Fifthly, this application proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned voice control method for industrial equipment based on a large language model.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] (1) The standardized operation instruction set constructed by the present invention standardizes the expression form of equipment operation instructions, realizes the structured conversion of voice instructions to operation instructions, solves the problem that the expression of voice instructions is not uniform and cannot be directly used for equipment control in the prior art, and improves the standardization and executability of operation instruction construction.

[0053] (2) The process knowledge retrieval method proposed in this invention, by integrating equipment process knowledge, supplements and constrains voice information, solves the problems of incomplete semantic information and difficulty in understanding professional terms in the prior art, significantly improves the professionalism and semantic understanding accuracy of voice control of industrial equipment, and improves the controllability and interpretability of model output.

[0054] (3) The reasoning method based on the thought chain proposed in this invention has the ability to stepwise reason and analyze complex instructions, solves the problem of difficult decomposition and reasoning of complex semantics in the prior art, improves the ability to understand the voice of users with multiple conditions and multiple intentions, and effectively improves the accuracy of understanding complex instructions and the reliability of execution. Attached Figure Description

[0055] Figure 1 This is a flowchart of a voice control method for industrial equipment based on a large language model, as described in an embodiment of the present invention.

[0056] Figure 2 This is a schematic diagram of a voice control method for industrial equipment based on a large language model, as described in an embodiment of the present invention.

[0057] Figure 3 This is a schematic diagram illustrating the construction of the operation instruction set in an embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram of the operation process in an embodiment of the present invention. Detailed Implementation

[0059] Example 1:

[0060] This invention provides a voice control method for industrial equipment based on a large language model, such as... Figure 1 and Figure 2 As shown, it includes the following steps:

[0061] Step 1: Analyze the voice control requirements of industrial equipment and construct a standardized set of operating instructions;

[0062] This paper analyzes the functional components and control methods of industrial equipment, clarifying the controllable objects, executable control actions, and conditional state information within the equipment. Based on this, a standardized operation instruction set is constructed to convert complex, diverse, and synonymous control intentions in human natural language into standardized structured instruction text. The standardized operation instruction set is constructed as follows: Figure 3 As shown, the specific steps include:

[0063] Step 1.1: Extract keywords from the voice control of industrial equipment and establish a keyword library for operation instructions;

[0064] Keywords are the basic elements that make up operating instructions and are closely related to the industrial equipment itself and the needs of users. Considering the characteristics of industrial equipment operation, keywords can be divided into three categories: controlled object keywords, execution action keywords, and conditional keywords. Corresponding thesauruses are then constructed to form the operating instruction keyword library.

[0065] (1) Controlled object thesaurus: This includes several keywords for controlled objects, representing the controlled objects (components) in industrial equipment, used to identify control targets from text information. For example: motors, flow valves, switching components, etc.

[0066] (2) Execution action vocabulary: This includes several execution action keywords, which are used to describe the specific control actions performed on the controlled objects in industrial equipment. These keywords are set as verbs related to control actions, clarifying the specific operational content of the instructions. For example: start, stop, open, close, etc.

[0067] (3) Conditional terminology: This includes several conditional keywords that describe the conditions for executing actions, such as delayed execution, conditional triggering, or equipment status query. Identifying these conditions enables flexible and complex control behaviors. For example: after how many minutes, when the temperature reaches the set value, when the pressure reaches the set value, query, current status, etc.

[0068] For example: User's voice: "I want to turn on the motor in 10 minutes." "Motor" is added to the controlled object vocabulary, "turn on" is added to the action vocabulary, and "in 10 minutes" is added to the condition vocabulary.

[0069] Step 1.2: Based on the functional matching relationship between the controlled object and the executed action, legally combine the keywords in the operation instruction keyword library to construct an operation instruction set; the operation instruction set includes several operation instructions, which include two types: basic operation instructions and conditional operation instructions;

[0070] (1) Construction of basic operation instructions: Based on the actual control functions of each controlled object, determine the type of action to be performed. Only when the action to be performed falls within the range of actions that the controlled object can perform, construct the corresponding basic operation instructions. The basic operation instructions are formed by combining the action keywords and the controlled object keywords. The structure is: action keyword + controlled object keyword.

[0071] For example, the user's voice command "Now turn on the motor" corresponds to the basic operation command: "Open—Motor". Here, "open" is the keyword for the action to be performed, and "motor" is the keyword for the controlled object.

[0072] (2) Construction of conditional operation instructions: Based on the basic operation instructions, a conditional dictionary is introduced to extend the execution methods of the basic operation instructions and construct conditional operation instructions that include delayed execution, conditional triggering, or status query. The structure of the conditional operation instructions is: conditional keyword + execution action keyword + controlled object keyword.

[0073] Step 1.3: Define each operation instruction in the operation instruction set in a structured form to obtain structured operation instructions, which are used to describe the operation of industrial equipment, and thus obtain a standardized operation instruction set;

[0074] The structured operation instructions include a controlled object field, an execution action field, and a condition field. The values ​​of the controlled object field, execution action field, and condition field are derived from corresponding dictionaries and constructed according to the operation instructions in the operation instruction set. If no condition keyword is included, the condition field is empty.

[0075] For example, if a user says "I want to turn on the motor in 10 minutes", the corresponding structured operation instruction generated is: delay 10 minutes — turn on — motor, where the controlled object field is motor, the action field is turn on, and the condition field is delay 10 minutes.

[0076] Step 2: Obtain process knowledge elements and construct a structured process knowledge set; the process knowledge elements are used to describe process knowledge related to industrial equipment;

[0077] First, analyze equipment control manuals, safety operating procedures, and equipment operation logs to extract process knowledge elements. Semantically decompose these process knowledge elements and divide them into multiple process knowledge units according to the principle of minimum expressible control meaning. Each process knowledge unit includes a knowledge identifier field, a knowledge content field, a knowledge type field, and a source field. The knowledge identifier field identifies different process knowledge units, the knowledge content field stores specific process knowledge content, the knowledge type field identifies the category to which the process knowledge content belongs (e.g., function, emergency), and the source field identifies the type of document or record from which the process knowledge content originates.

[0078] Then, the various process knowledge units are organized in a structured manner to form a structured process knowledge set for subsequent semantic retrieval processing.

[0079] Step 3: Based on the standardized operation instruction set and structured process knowledge set, the user's voice input signal is converted into standardized operation instructions using a large language model and then used to drive industrial equipment to perform actions;

[0080] like Figure 4 As shown, it includes the following steps:

[0081] Step 3.1: Obtain the raw voice signal input by the user and convert it into text information through speech recognition;

[0082] In this embodiment, the input raw speech signal is first preprocessed by noise reduction and voice activity detection (VAD) to obtain valid audio segments. Secondly, neural networks and end-to-end technologies are used to model and sequence the valid audio segments, directly converting them into corresponding text information. Finally, the identified text information is post-processed, including text normalization, punctuation completion, and simple error correction, to output the final text information.

[0083] Step 3.2: Retrieve process knowledge units associated with text information from the structured process knowledge set to form process knowledge information;

[0084] First, the text information is preprocessed and keywords are extracted: the text information is segmented, stop words are removed, and keywords related to equipment control are retained. Second, the extracted keywords are used to form a retrieval matching feature set. At the same time, the knowledge content field and knowledge type field of each process knowledge unit are read from the structured process knowledge set to form a set of knowledge to be matched. Then, the matching degree (relevance) of the retrieval matching feature set and each process knowledge unit in the set of knowledge to be matched is calculated. At the same time, an initial matching priority is set for each process knowledge unit. When the large language model determines that the knowledge content field of the process knowledge unit contains a description that is consistent with or semantically similar to the keywords of the controlled object or the keywords of the execution action, the matching priority of the process knowledge unit is increased. When the large language model determines that the knowledge type field of the process knowledge unit matches the intent type of the text information, its matching priority is increased. Process knowledge units that meet the preset screening conditions are selected to form process knowledge information composed of several process knowledge units. The original text information obtained from user speech recognition is retained for subsequent model prompt word construction.

[0085] The method for selecting process knowledge units that meet the preset screening conditions is as follows:

[0086] First, select according to matching priority; when matching priorities are the same, select according to matching degree.

[0087] In this embodiment, the text information is: "I want to turn on the motor in 10 minutes." Two candidate process knowledge units were retrieved: K01: content "The motor supports delayed start control mode", knowledge type field is function type; K02: content "Delay control can be used for timed start-up operation of equipment", knowledge type field is function type; K03: content "When the motor has a fault alarm, the emergency stop button should be pressed immediately to cut off the power", knowledge type field is emergency type. It is clear that the function type matches the intent type of the text information, therefore, the matching priority of K01 and K02 is higher than that of K03.

[0088] Step 3.3: Based on the preset thought chain template, guide the large language model to reason according to text information and process knowledge information, output operation instructions, and drive industrial equipment to perform actions;

[0089] In this embodiment, a thought chain reasoning mechanism is used to guide the large language model to reason and analyze the enhanced semantic description composed of text information and process knowledge information. It identifies the controlled object, the execution action and the conditions in sequence, infers its control intention, and matches the identification results with the operation instruction set to finally output the corresponding equipment operation instructions.

[0090] Step 3.3.1: Construct model prompt words based on textual information and process knowledge information;

[0091] First, textual information is used as the basic semantic input. Process knowledge information is introduced to supplement and constrain the textual semantics. The two are then combined to provide background information related to industrial equipment control, forming an enhanced semantic description. Second, the enhanced semantic description is structured to obtain model prompts that include both the original textual information input by the user and the process knowledge information.

[0092] Step 3.3.2: Perform multi-intent recognition on the model prompts and divide the model prompts into several independent statements according to the recognized control intent;

[0093] Specifically: determine whether the original speech signal contains multiple independent control intentions; when the original speech signal contains multiple control intentions, split the model prompt words into several independent sentences according to the sentence order, and perform subsequent analysis on each sentence.

[0094] Step 3.3.3: For each independent statement, identify the keywords of the controlled object, and take the industrial equipment or component pointed to by the identified keywords of the controlled object as the control target;

[0095] Step 3.3.4: For each individual statement, identify the keywords of the execution action to clarify the specific control behavior;

[0096] Step 3.3.5: For each independent statement, identify condition keywords to characterize the execution conditions of the control behavior, including delayed execution conditions, condition triggering, and status query conditions;

[0097] Step 3.3.6: Integrate all identified keywords and determine the validity of the control intent. If no controlled object keyword or execution action keyword is identified, it is considered an invalid control intent; if at least the controlled object keyword and execution action keyword are identified, it is considered a valid control intent.

[0098] For example, the enhanced semantic description is: "Stop the motor now, and turn it back on in 10 minutes." The inference analysis identifies two control intentions: control intention... Figure 1 The controlled object is a motor, and the action to be performed is to stop; the control intention is... Figure 2 The controlled object is a motor, the action to be performed is to turn it on, and the condition is after 10 minutes.

[0099] Step 3.3.7: After confirming the effectiveness of the control intent, the extracted keywords are matched with the standardized operation instruction set to obtain the standardized operation instructions;

[0100] The specific processing method is as follows:

[0101] (1) Determining the type of operation instruction: Based on the extracted keywords, determine the type of operation instruction to be matched: when the keywords only contain the keywords of the controlled object and the keywords of the execution action, determine the basic operation instruction to be matched; when the keywords further contain conditional keywords, determine the conditional operation instruction to be matched.

[0102] (2) Operation instruction field mapping: Controlled object keywords are mapped to controlled object fields, execution action keywords are mapped to execution action fields, and condition keywords are mapped to condition fields, forming standardized operation instruction matching items.

[0103] (3) Operation instruction search and determination: In the standardized operation instruction set, search for operation instructions that have the same field values ​​as the standardized operation instruction matching item. When there is a unique matching operation instruction, determine that operation instruction as the target operation instruction; when no matching operation instruction is found, the output command is invalid.

[0104] For example, the command "Delay for 10 minutes to turn on the motor" is processed and analyzed into a single control intent. The resulting keyword mapping is: the controlled object field is "motor", the action field is "turn on", and the condition field is "delay for 10 minutes". Based on this, the corresponding operation instruction is determined in the standardized operation instruction set.

[0105] Step 3.3.8: Output standardized operating instructions to drive industrial equipment to perform actions;

[0106] Standardized operating instructions are converted into a format recognizable by the equipment and mapped to corresponding control signals. These control signals are transmitted to the execution terminal via a pre-defined industrial communication interface, driving the target industrial equipment to complete the corresponding physical actions, thus achieving voice control of the industrial equipment. The execution terminal includes, but is not limited to, programmable logic controllers, embedded microcontroller systems, or other industrial control units, thereby enabling the control of the industrial equipment.

[0107] Example 2:

[0108] A voice control system for industrial equipment based on a large language model, used to implement a voice control method for industrial equipment based on a large language model, includes:

[0109] The instruction set construction module is used to analyze the voice control requirements of industrial equipment and build a standardized operation instruction set.

[0110] The process knowledge module is used to acquire process knowledge elements and construct a structured process knowledge set.

[0111] The retrieval and control module is used to convert user-input voice signals into standardized operating instructions based on a standardized set of operating instructions and a structured set of process knowledge, and to drive industrial equipment to perform actions using a large language model.

[0112] Example 3:

[0113] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the aforementioned industrial equipment voice control method based on a large language model.

[0114] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements a voice control method for industrial equipment based on a large language model as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0115] The processor is used to execute all or part of the steps in the industrial equipment voice control method based on a large language model as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0116] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the industrial equipment voice control method based on a large language model described in the above embodiments.

[0117] Example 4:

[0118] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0119] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the industrial equipment voice control method based on a large language model described in various embodiments of this application.

[0120] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) application stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the aforementioned industrial equipment voice control method based on a large language model.

[0121] Example 5:

[0122] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned industrial equipment voice control method based on a large language model.

[0123] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0124] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0125] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. A voice control method for industrial equipment based on a large language model, characterized in that, Includes the following steps: Analyze the voice control requirements of industrial equipment and construct a standardized set of operating instructions; Acquire process knowledge elements and construct a structured process knowledge set; the process knowledge elements are used to describe process knowledge related to industrial equipment; Based on a standardized set of operating instructions and a structured set of process knowledge, a large language model is used to convert user-input voice signals into standardized operating instructions and drive industrial equipment to perform actions. This includes the following steps: B1: Acquire the user's original voice signal, perform speech recognition, and convert it into text information; B2: Retrieve process knowledge units associated with text information from the structured process knowledge set to form process knowledge information; B3: Based on a preset thought chain template, the large language model is guided to reason based on textual information and process knowledge information, output operation instructions, and drive industrial equipment to perform actions; The B3 specifically includes the following steps: B3.1: Construct model prompt words based on textual information and process knowledge information; First, textual information is concatenated with process knowledge information to form an enhanced semantic description; then, the enhanced semantic description is structured and organized to obtain model prompt words. B3.2: Perform multi-intent recognition on model prompts and divide the model prompts into several independent statements according to the recognized control intent; Specifically: determine whether the original speech signal contains multiple independent control intentions; when the original speech signal contains multiple control intentions, split the model prompt words into several independent sentences according to the sentence order; B3.3: For each independent statement, identify the keywords of the controlled object, and take the industrial equipment pointed to by the identified keywords of the controlled object as the control target; B3.4: For each individual statement, identify the keywords for the execution action; B3.5: For each individual statement, identify conditional keywords; B3.6: Integrate all identified keywords and determine the validity of the control intent. If neither the controlled object keyword nor the execution action keyword is identified, the control intent is deemed invalid. If at least the controlled object keyword and the execution action keyword are identified, the control intent is considered valid. B3.7: After confirming the validity of the control intent, the extracted keywords are matched with the standardized operating instruction set to obtain the standardized operating instructions; The specific processing method is as follows: (1) Determining the type of operation instruction: Based on the extracted keywords, determine the type of operation instruction to be matched: when the keywords only contain keywords of the controlled object and keywords of the action to be executed, determine the basic operation instruction to be matched; when the keywords further contain conditional keywords, determine the conditional operation instruction to be matched. (2) Operation instruction field mapping: Controlled object keywords are mapped to controlled object fields, execution action keywords are mapped to execution action fields, and condition keywords are mapped to condition fields, forming standardized operation instruction matching items; (3) Operation instruction search and determination: In the standardized operation instruction set, search for operation instructions that have the same field values ​​as the standardized operation instruction matching items. When there is a unique matching operation instruction, determine that operation instruction as the target operation instruction; when no matching operation instruction is found, the output command is invalid. B3.8: Outputs standardized operating instructions to drive industrial equipment to perform actions.

2. The voice control method for industrial equipment based on a large language model according to claim 1, characterized in that, The analysis of the voice control requirements of industrial equipment and the construction of a standardized operation instruction set specifically include: A1: Extract keywords for voice control of industrial equipment and establish a keyword library for operation instructions; Keywords are categorized into three types: controlled object keywords, execution action keywords, and conditional keywords. Corresponding thesauruses are then constructed to collectively form the operation instruction keyword library. (1) Controlled object thesaurus: This includes several controlled object keywords, representing the controlled objects in industrial equipment; (2) Execution action vocabulary: This includes several execution action keywords, which are used to describe the specific control actions performed on the controlled objects in industrial equipment; (3) Conditional terminology: includes several conditional keywords that describe the conditions for performing an action; A2: Legally combine keywords from the operation instruction keyword library to construct an operation instruction set; the operation instruction set includes several operation instructions, which include two types: basic operation instructions and conditional operation instructions. (1) The basic operation instructions are formed by combining the execution action keywords and the controlled object keywords. The structure is execution action keywords + controlled object keywords; (2) Conditional operation instructions are based on basic operation instructions, with the introduction of a conditional word library to extend the execution method of basic operation instructions. The structure of conditional operation instructions is conditional keyword + execution action keyword + controlled object keyword; A3: Define each operation instruction in the operation instruction set in a structured form to obtain structured operation instructions, and then obtain a standardized operation instruction set; The structured operation instructions include a controlled object field, an execution action field, and a condition field. The values ​​of the controlled object field, the execution action field, and the condition field are derived from the corresponding thesaurus and constructed according to the operation instructions in the operation instruction set. When no condition keyword is included, the condition field is empty.

3. The voice control method for industrial equipment based on a large language model according to claim 1, characterized in that, The acquisition of process knowledge elements and the construction of a structured process knowledge set specifically involves: First, process knowledge elements are extracted and semantically decomposed into multiple process knowledge units. Each process knowledge unit includes a knowledge identifier field, a knowledge content field, a knowledge type field, and a source field. The knowledge identifier field is used to identify different process knowledge units, the knowledge content field is used to store specific process knowledge content, the knowledge type field is used to identify the category to which the process knowledge content belongs, and the source field is used to identify the type of document or record from which the process knowledge content originates. Then, the process knowledge units are organized in a structured manner to form a structured process knowledge set.

4. A voice control system for industrial equipment based on a large language model, used to implement the voice control method for industrial equipment based on a large language model as described in any one of claims 1-3, characterized in that, include: The instruction set construction module is used to analyze the voice control requirements of industrial equipment and build a standardized operation instruction set. The process knowledge module is used to acquire process knowledge elements and construct a structured process knowledge set. The retrieval and control module is used to convert user-input voice signals into standardized operating instructions based on a standardized set of operating instructions and a structured set of process knowledge, and to drive industrial equipment to perform actions using a large language model.

5. An electronic device, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the industrial equipment voice control method based on a large language model as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed, cause the processor to perform the industrial equipment voice control method based on a large language model as described in any one of claims 1-3.

7. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the voice control method for industrial equipment based on a large language model as described in any one of claims 1-3.

Images