Injection molding machine process experience sedimentation and intelligent auxiliary system based on voice interaction

Abstract

The application relates to an injection molding machine process experience precipitation and intelligent auxiliary system based on voice interaction, relates to the field of injection molding interaction control, and comprises a voice understanding engine in the field of injection molding, which is used for collecting voice information under an injection molding process scene and accurately analyzing voice instructions and selection instructions; a context perception question and answer module, which is used for collecting a process state in response to the voice instructions and generating a query vector according to the process state and the voice instructions; a knowledge graph construction engine, which is used for forming a searchable experience knowledge base, traversing an associated path from the experience knowledge base and determining a candidate scheme set according to path similarity; an operation interaction module, which is used for displaying the candidate scheme set and generating a parameter modification instruction in response to the selection instruction; and an instruction execution module, which is used for converting the parameter modification instruction into a parameter modification or a function call executable by a control system, and completing real-time operation on the injection molding machine. The application has the effect of improving the convenience of injection molding interaction control.

Description

Technical Field

[0001] This invention relates to the field of injection molding interactive control, and in particular to an injection molding machine process experience accumulation and intelligent assistance system based on voice interaction. Background Technology

[0002] Injection molding machines, as core production equipment widely used in daily life, industrial manufacturing, and even aerospace, are gradually incorporating multimodal biometric technologies such as facial recognition and voice interaction to address practical needs like access control and operational optimization. However, the application of voice interaction technology remains at a superficial stage, with almost no effective application scenarios observed, and machine-human interaction has not yet achieved proactivity and personalization. This situation clearly lags significantly behind the future development direction of industrial interaction.

[0003] Currently, the application of voice technology in injection molding machines is mainly focused on basic operations such as page switching or parameter adjustment via commands. While these applications alleviate some of the problems of cumbersome multi-level menu navigation and inconvenient operation in busy scenarios associated with traditional HMIs, their limitations are also obvious: On the one hand, faced with complex and diverse process parameters, operational experience remains at the individual level, making it difficult to record and pass on tacit knowledge through voice interaction. On the other hand, injection molding machines are always in a passive execution state, only able to respond to commands but unable to engage in any form of positive interaction with the operator. This simple "command-execution" mode lacks both an understanding of the process context and the ability to utilize historical experience for auxiliary judgment, making it difficult for voice technology to truly integrate into the core scenarios of injection molding production. Summary of the Invention

[0004] To improve the convenience of interactive control in injection molding, this invention provides a voice-interactive injection molding machine process experience accumulation and intelligent assistance system.

[0005] This invention provides a voice-interactive injection molding machine process experience accumulation and intelligent assistance system, employing the following technical solution: A voice-interactive injection molding machine process experience accumulation and intelligent assistance system includes: A voice understanding engine for the injection molding field, used to collect voice information in injection molding process scenarios and accurately parse voice commands and selection commands; The context-aware question-answering module is used to collect process status in response to voice commands and generate query vectors based on process status and voice commands. The knowledge graph building engine is used to structure the preset operation logs to form a searchable experience knowledge base, and in response to the query vector, it traverses the associated paths from the experience knowledge base and calculates the path similarity, and determines the set of alternative solutions based on the path similarity. The operation interaction module is used to display a set of alternative solutions and generate parameter modification instructions in response to selection instructions; The instruction execution module is used to convert parameter modification instructions into parameter modifications or function calls that can be executed by the control system, thereby completing the real-time operation of the injection molding machine.

[0006] Optionally, the speech understanding engine for the injection molding field includes: Determine the voice intent based on the voice command; When the voice intent is inconsistent with the preset fault diagnosis intent, the voice command is sent to the command execution module.

[0007] Optional, also includes: The operation history module is used to retrieve control status and execution results in response to parameter modification commands; An operation log is generated by combining voice commands, control status, and execution results.

[0008] Optionally, the speech understanding engine for the injection molding field includes a dedicated injection molding lexicon and a parameter mapping table; The data sources for the injection molding-specific terminology database include injection molding machine operation manuals, historical fault repair records, real operation dialogues collected on-site, and professional terminology. The parameter mapping table includes a preset mapping library and user-defined extensions.

[0009] By adopting the above technical solutions, this system upgrades the injection molding machine from passive execution to context-aware assistance through voice parsing, intent distribution, knowledge graph retrieval, and operation log closed loop. Intent distribution reduces unnecessary retrieval of non-fault instructions, operation logs support dynamic accumulation of experience, and dedicated lexicon and parameter mapping ensure accurate recognition of injection molding terminology, thereby improving the convenience of injection molding interactive control.

[0010] In summary, this application includes at least one of the following beneficial technical effects: Through a collaborative architecture that integrates a voice understanding engine, a context-aware question-and-answer module, a knowledge graph construction engine, an operation interaction module, and an instruction execution module for injection molding, voice-driven process assistance and experience accumulation for injection molding machines are achieved. This reduces the problem of traditional voice control being merely passive and unable to provide intelligent decision-making in conjunction with process scenarios, thereby improving the convenience of injection molding interactive control.

[0011] Voice intent classification is used to divert commands. Fault diagnosis commands are sent to the question and answer module, while other commands are directly connected to the execution module, reducing invalid searches and process redundancy, and improving the efficiency of voice control response.

[0012] By synchronously recording voice commands, control status, and execution results to form a complete operation log, a reliable data source is provided for the knowledge graph, ensuring the continuous iteration and updating of the experience knowledge base. Attached Figure Description

[0013] Figure 1 This is a system block diagram of an injection molding machine process experience accumulation and intelligent auxiliary system based on voice interaction.

[0014] The parts referred to by the numbers in the above attached diagrams are as follows: 1. Speech understanding engine for injection molding; 2. Context-aware question answering module; 3. Knowledge graph construction engine; 4. Operation interaction module; 5. Command execution module; 6. Operation history module. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] This application discloses a voice-interactive injection molding machine process experience accumulation and intelligent assistance system.

[0017] Reference Figure 1 Based on voice interaction, the injection molding machine process experience accumulation and intelligent assistance system includes: The injection molding field voice understanding engine 1 is used to collect voice information in the injection molding process scenario, accurately parse the voice commands and selection commands, and determine the voice intent based on the voice commands.

[0018] When the voice intent is inconsistent with the preset fault diagnosis intent, the voice command is sent to the command execution module 5.

[0019] The speech understanding engine 1 for the injection molding field includes a dedicated injection molding lexicon and a parameter mapping table.

[0020] The data sources for the injection molding-specific terminology database include injection molding machine operation manuals, historical fault repair records, real-world operation dialogues collected on-site, and professional terminology.

[0021] The parameter mapping table includes a preset mapping library and user-defined extensions.

[0022] Voice information refers to the raw voice produced by the operator, which can be collected using an industrial-grade microphone array.

[0023] Voice commands refer to the text results obtained by recognizing and parsing voice information. A voice recognition model can be jointly trained by fusing a general dataset and a special lexicon for injection molding, and a parameter mapping table can be established to achieve accurate parsing of voice commands in injection molding process scenarios.

[0024] The injection molding-specific lexicon data sources include injection molding industry-specific corpora and professional terminology-enhanced corpora. The injection molding industry-specific corpora include: injection molding machine operation manual texts, historical fault repair records, and real operation dialogues collected on-site (with different accents and different noise environments), etc.; the professional terminology-enhanced corpora resample and enhance injection molding-specific terms such as "flash," "weld line," "ejector pin," and "pressure holding" to improve the recognition rate of rare words.

[0025] The parameter mapping table construction includes a preset mapping library and user-defined extensions. The preset mapping library contains built-in mapping relationships between commonly used process parameters and PLC variables; the user-defined extensions provide a visual configuration interface, allowing engineers to associate and bind custom variable names with natural language descriptions in voice commands.

[0026] For example, when an operator issues the voice message "Switch to homepage", the injection molding field speech understanding engine 1 first converts the speech into a phoneme sequence through an acoustic model, and then decodes it into the text "Switch to homepage" through a language model combined with an injection molding-specific dictionary. The keywords "switch" and "homepage" are extracted, and the parameter mapping table is based on predefined mapping rules. "Switch" corresponds to the page navigation function in the control system, and "homepage" maps to the corresponding page address, thus converting the speech information into a voice command composed of the page navigation function and the page address.

[0027] A selection command is an instruction issued by the operator when selecting a parameter modification scheme, such as "select the first scheme". The method for determining the selection command is the same as the above-mentioned voice commands.

[0028] Voice intent refers to the intended operation of a voice command. Figure 1 Generally, voice commands include three categories: parameter modification, function call, and fault diagnosis. If the voice intent differs from the fault diagnosis intent, it indicates that the voice intent pertains to parameter modification or function call, meaning the operator has issued a clear operational intent. In this case, the voice command can be directly sent to the command execution module 5 for execution. Figure 1 If the operator does not issue a specific operational intention, then the specific parameter modification plan needs to be generated by the context-aware question-and-answer module 2 and other modules, and then sent to the instruction execution module 5 for execution.

[0029] For example, when an operator sends a voice message: "Switch to homepage", the voice intent belongs to the function call category. When an operator sends a voice message: "The product has been downgraded, what should I do?", the voice intent belongs to the fault diagnosis category.

[0030] Context-aware question answering module 2 is used to collect process status in response to voice commands and generate query vectors based on process status and voice commands.

[0031] Process status refers to parameters of the current production situation, such as materials, mold, current injection molding parameters, and stage.

[0032] A query vector is a multidimensional array obtained by transforming the process status and used to query specific parameter modification schemes. The query vector contains three dimensions: problem dimension (phenomenon description), process dimension (process conditions), and environment dimension (environmental factors).

[0033] For example, when an operator sends a voice message: "The product has shrunk, what should I do?", the query vector is: Problem dimension: Fault phenomenon "shrinkage"; Process dimension: Current material "ABS", wall thickness "3mm", holding pressure "85MPa", current stage "holding pressure"; Environment dimension: Mold ID "M-203", ambient temperature "26℃".

[0034] The knowledge graph construction engine 3 is used to structure the preset operation logs to form a searchable experience knowledge base, and in response to the query vector, it traverses the associated paths from the experience knowledge base and calculates the path similarity, and determines the set of alternative solutions based on the path similarity.

[0035] An operation log is a collection of information that records each voice command and the context of the injection molding machine's process status. It can retrieve historical operation records, alarm events, and solutions as operation logs. These historical operation records are not limited to voice operation history records.

[0036] The experience knowledge base refers to a knowledge graph formed by structuring historical operation records, alarm events, and solutions. The smallest unit of the knowledge graph building engine 3 is the process case, and its node definition and relationship modeling are as follows: 1. Node type design: Fault phenomenon node: such as "shrinkage", "flash", etc.; Solution node: contains specific parameter adjustment records and operation steps; Process condition node: material, mold, parameter range; Operator node: experienced engineer identifier; 2. Relationship definition: Phenomenon - Occurrence - Condition; Phenomenon - Solution - Operator; Solution - Effect - Success rate.

[0037] Path similarity is a numerical value used to evaluate the applicability of parameter modification schemes. When the knowledge graph construction engine 3 receives the query vector sent by the context-aware question-answering module 2, it uses a graph neural network (GNN) to calculate similarity, traversing all associated paths of a fault phenomenon node and calculating the matching degree between each path and the query vector. Because process knowledge naturally has a graph structure (multiple associations between phenomenon-solution-condition-operator), GNN can effectively capture this complex relationship, making it more accurate than traditional vector retrieval. The path similarity calculation formula is: Sim = K1 * text sim +K2*param sim +K3*context sim Where Sim is the path similarity, and text sim For the question dimension, param sim From a process perspective, context sim For the environmental dimension, K1, K2, and K3 are adjustable weight coefficients.

[0038] The alternative solution set refers to the set of parameter modification solutions that are most suitable for the current production situation, selected based on path similarity. Generally, the three historical cases with the highest path similarity are selected and returned in descending order of similarity as the alternative solution set.

[0039] Operation interaction module 4 is used to display a set of alternative solutions and generate parameter modification instructions in response to selection instructions.

[0040] The parameter modification command refers to the set of specific parameter modifications generated according to the parameter modification scheme in the set of alternative schemes selected by the operator. The operation interaction module 4 displays the details of the alternative scheme set through voice synthesis broadcast or HMI interface. The operator can issue a voice command to form a selection command or select the target scheme as the selection command through HMI touch. When key parameters are modified, the system will initiate a secondary confirmation, and the operation can only be executed after the operator confirms.

[0041] For example, when the operator issues a voice message: "The product has shrunk, what should I do?", the operation interaction module 4 responds through voice synthesis or HMI interface: "According to historical records, when dealing with shrinkage issues of ABS material with a wall thickness of 3mm, there have been two effective solutions: Solution A, increase the holding pressure by 12%; Solution B, extend the holding time by 2 seconds. Your current holding pressure is 85MPa. We suggest trying Solution A. Do you want to proceed?" The operator selects "Execution Plan A" via voice command or HMI touch control. The system then initiates a secondary confirmation: "Confirm change of holding pressure?". After the operator confirms, a signal is sent to the command execution module 5.

[0042] The instruction execution module 5 is used to convert parameter modification instructions into parameter modifications or function calls that can be executed by the control system, thereby completing the real-time operation of the injection molding machine.

[0043] The parsed instructions are converted into parameter modifications or function calls that the control system can execute, enabling real-time operation of the injection molding machine. Parameter range verification is performed during execution, and results are fed back after execution.

[0044] For example, after the operator selects "Execution Plan A" via voice command or HMI touch, the command execution module 5 will increase the holding pressure by 12%, that is, modify it to 95.2MPa.

[0045] Operation history module 6 is used to retrieve control status and execution results in response to parameter modification commands.

[0046] An operation log is generated by combining voice commands, control status, and execution results.

[0047] The operation history module 6 records each voice command and the injection molding machine process status context to form a traceable operation log, providing a data foundation for the subsequent construction of an experience knowledge base.

[0048] The single operation history record is the smallest unit in this module, and it consists of the following: 1. Voice command layer: Recognized text, parsed keyword set, and original voice file (optional); 2. Process Status Layer: Synchronously records snapshots of the control system status when commands are issued, including: a. Current process status (injection / holding pressure / cooling / ejection, etc.); b. Real-time values ​​of key parameters (pressure, speed, temperature, position, etc.); c. Alarm status (whether there is an active alarm); d. Mold ID, material grade, etc. 3. Operation Result Layer: Records the command execution result (success / failure), parameter changes after execution, and subsequent equipment operating status (whether a new alarm is generated).

[0049] For example, the operation history for this operation consists of: Voice command layer: "Product size reduced," "Increase holding pressure by 12%" Process status layer: Current process status (ejection), current holding pressure is 85MPa, alarm status (whether the alarm is activated), ABS material, wall thickness 3mm; Operation result layer: The command was executed successfully, the holding pressure was adjusted to 95.2MPa, and there was no alarm.

[0050] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0051] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A voice-interactive injection molding machine process experience accumulation and intelligent assistance system, characterized in that, include: A voice understanding engine for the injection molding field (1) is used to collect voice information in the injection molding process scenario and accurately parse out voice commands and selection commands; The context-aware question answering module (2) is used to collect the process status in response to voice commands and generate a query vector based on the process status and voice commands; The knowledge graph construction engine (3) is used to structure the preset operation logs to form a searchable experience knowledge base, and to traverse the associated paths from the experience knowledge base and calculate the path similarity in response to the query vector, and to determine the set of alternative solutions based on the path similarity. The operation interaction module (4) is used to display the set of alternative solutions and generate parameter modification instructions in response to the selection instructions; The instruction execution module (5) is used to convert parameter modification instructions into parameter modification or function calls that can be executed by the control system, so as to complete the real-time operation of the injection molding machine.

2. The voice-interactive injection molding machine process experience accumulation and intelligent assistance system according to claim 1, characterized in that, The speech understanding engine (1) for the injection molding field includes: Determine the voice intent based on the voice command; When the voice intent is inconsistent with the preset fault diagnosis intent, the voice command is sent to the command execution module (5).

3. The voice-interactive injection molding machine process experience accumulation and intelligent assistance system according to claim 1, characterized in that, Also includes: The operation history module (6) is used to retrieve the control status and execution results in response to parameter modification commands; An operation log is generated by combining voice commands, control status, and execution results.

4. The voice-interactive injection molding machine process experience accumulation and intelligent assistance system according to claim 1, characterized in that, The speech understanding engine (1) for the injection molding field includes a dedicated injection molding lexicon and a parameter mapping table; The data sources for the injection molding-specific terminology database include injection molding machine operation manuals, historical fault repair records, real operation dialogues collected on-site, and professional terminology. The parameter mapping table includes a preset mapping library and user-defined extensions.

Images