Voice control based industrial horcrux workbench intelligent management method

Abstract

The application discloses a speech control-based intelligent management method for an industrial horcrux workstation, and belongs to the technical field of industrial control. The method comprises the following steps: collecting operator speech audio, and parsing the speech audio into device operation text instructions after permission is passed; establishing a data communication link with a production management system, downloading process operation guidance data, and generating a cooperative control instruction set; issuing the cooperative control instruction set to a communication interface of an industrial horcrux controller, and synchronously acquiring real-time running parameters of a target workstation device through the communication interface; and matching the real-time running parameters with device state constraint conditions. The application solves the problem of disconnection between process execution and device state and low human-computer interaction efficiency leading to interruption of production rhythm in the prior art through real-time comparison of speech instructions and process requirements, continuous verification of device state and automatic generation of a whole-process log.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, and in particular to a voice-controlled intelligent management method for industrial HarmonyOS workbenches. Background Technology

[0002] As industrial manufacturing rapidly evolves towards digitalization and intelligence, production sites are placing higher demands on human-machine interaction efficiency, process execution accuracy, and process traceability. Traditional workbenches are gradually revealing their shortcomings in areas such as equipment collaboration, process control, and data recording.

[0003] In existing technologies, some information-based workstations are equipped with industrial computers or tablets to display electronic work instructions and input basic data. However, this approach still has significant drawbacks: Firstly, there is a lack of a unified collaborative control mechanism between equipment, resulting in a disconnect between process execution and equipment status. Operators can continue working even when equipment does not meet process requirements, leading to frequent hidden quality violations. Secondly, the human-machine interaction method is limited, making it difficult for operators to operate the system conveniently when their hands are occupied by tools, frequently interrupting work and affecting production rhythm. Furthermore, production process data largely relies on manual entry, making it difficult to trace specific steps, operators, and equipment status when quality abnormalities occur, resulting in long troubleshooting cycles and recurring similar problems. Summary of the Invention

[0004] This application provides a voice-controlled industrial HarmonyOS workbench intelligent management method, which solves the problems in the prior art such as the disconnect between process execution and equipment status, low human-machine interaction efficiency leading to production cycle interruption, and difficulty in tracing specific work steps and operators after anomalies occur. It realizes intelligent production management from source error prevention and closed-loop process control to accurate post-event traceability.

[0005] This application provides a voice-controlled intelligent management method for industrial HarmonyOS workbench, including: collecting operator voice audio, confirming operation permissions and extracting operator identity identifiers through voiceprint comparison, and parsing the voice audio into device operation text commands after the permissions are granted. Establish a data communication link with the production management system, download process operation guidance data, extract the equipment status constraints of the current process step, compare the equipment operation text instructions with the equipment status constraints, and generate a collaborative control instruction set. The collaborative control instruction set is sent to the communication interface of the industrial HarmonyOS controller to trigger the linkage operation of multiple target workbench devices, and the real-time operating parameters of the target workbench devices are synchronously obtained through the communication interface. The system matches real-time operating parameters with equipment status constraints. If the match fails, a voice alarm command is triggered. If the match is successful, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system.

[0006] Furthermore, the steps of confirming operation permissions and extracting operator identification through voiceprint comparison, and then parsing the voice audio into device operation text commands after permission is granted, include: Raw speech audio, including background noise, was captured using an industrial microphone; Silence detection and frame segmentation are performed on the original speech audio, the acoustic feature vectors of each speech frame are extracted, and the first feature matrix is ​​constructed by concatenating them. Read the authorized personnel voiceprint reference matrix library from the system memory, and calculate the similarity distance between the first feature matrix and each matrix in the voiceprint reference matrix library; When the minimum similarity distance is less than the safety threshold, the operation permission is approved, and the corresponding operator's identity is extracted. With permissions granted, the raw voice audio input to the acoustic decoding module is mapped to a Chinese string; Perform word segmentation and part-of-speech tagging on Chinese strings, remove stop words and retain action verbs and target nouns, convert the retained action verbs and target nouns into equipment operation text instructions, and synchronously load the equipment operation text instructions and operator identification into the internal data bus for later use.

[0007] Furthermore, the step of comparing equipment operation text commands with equipment status constraints to generate a collaborative control command set includes: Download the process operation instruction data file for the current product batch; Parse the work instruction data file and construct a directed acyclic graph containing multiple work step nodes; Obtain the current execution timestamp, locate the corresponding current step node in the directed acyclic graph, and read the device type and working parameter constraint range associated with the node; Match the target nouns in the device operation text instructions in the data bus with the device type; When the characters match, the action verb is converted into a low-level control primitive, and supplementary control primitives for linking external monitoring equipment are generated based on the working parameter constraint range. Align and merge the timelines of the underlying control primitives and the supplementary control primitives to generate a collaborative control instruction set.

[0008] Furthermore, the real-time operating parameters of the target workbench device are synchronously acquired through the communication interface: Disassemble the collaborative control instruction set and separate the master control communication data packet and the linkage communication data packet; Through kernel mechanisms, master control communication data packets are routed to the general asynchronous transceiver interface and sent to the master operating device to execute physical actions; Simultaneously, the linkage communication data packets are routed to the local area network bus interface and sent to the auxiliary equipment to initiate the response; during the linkage operation cycle, the data polling process is started; Configure a fixed sampling interval and read the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface; The analog voltage signal is converted into a digital quantized value, and the digital quantized values ​​at the same sampling time are concatenated into a multi-dimensional data vector, which is then stored in the cache as a real-time operating parameter.

[0009] Furthermore, if the matching fails, a voice alarm command is triggered; if the matching succeeds, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system. This process includes: Extract the multidimensional data vector from the cache and use a sliding time window to extract continuous parameter subsequences; Compare the mean of the parameter subsequence with the device state constraint interval; If the mean exceeds the constraint range, a hardware interrupt signal is generated, driving the speaker module to play a voice alarm audio containing the name of the out-of-bounds device; If the mean is within the constraint range, then extract the timestamp, multidimensional data vector, operation text instructions, and operator identification. The tamper-proof verification code of the extracted data is calculated using a hash algorithm; The extracted data and verification code are encapsulated into a standard message in Extensible Markup Language format and pushed to the production management system for archiving records.

[0010] Furthermore, the steps of performing silence detection and frame segmentation on the original speech audio, extracting the acoustic feature vectors of each speech frame, and concatenating them to construct the first feature matrix include: Based on preset window length and window shift parameters, a Hamming window function is applied to the original speech audio to perform frame segmentation, suppress spectral leakage, and obtain multiple overlapping audio short frames. For each audio frame, a fast Fourier transform algorithm is performed to convert the time-domain signal into frequency-domain energy distribution spectrum data. The frequency domain energy distribution spectrum data is passed through a Mel scale filter bank containing multiple triangular filters to calculate the absolute value of the logarithmic energy in each frequency band. Perform a discrete cosine transform on the resulting logarithmic energy absolute value sequence to eliminate the correlation within the data matrix and retain the quantized data of the preceding low-frequency dimensions; All short frames of low-frequency dimensional quantized data are sequentially stitched together in a row-level fashion to generate a two-dimensional numerical array containing the dynamic evolution of acoustics. This array is defined as the first feature matrix for underlying alignment.

[0011] Furthermore, the steps for parsing the process instruction data file and constructing a directed acyclic graph containing multiple process step nodes include: Extract the Extensible Markup Language (XML) document structure from the process instruction data file; Traverse the process level tags in the document structure and extract the unique identifier, the set of prerequisite dependency identifiers, and the process parameter thresholds for each operation. Initialize a blank graph data structure in computer memory; Instantiate the unique identifier of each operation as an independent graph node in the graph data structure; Read the set of node prerequisite dependency identifiers and generate connection edges with directional attributes between source graph nodes and target graph nodes based on the dependency relationships. Attach the device type and working parameter constraint range attribute fields to each independent graph node; Perform a depth-first traversal algorithm on the generated graph data structure to verify acyclic logic; If no closed-loop path is detected, a directed acyclic graph containing multiple work step nodes is output to the system cache.

[0012] Furthermore, the steps of configuring a fixed sampling interval and reading the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface include: Read the current count value of the high-precision hardware timer in the industrial HarmonyOS controller kernel and set the polling reference time based on microseconds; Register parallel read tasks for the general asynchronous transceiver interface and the local area network bus interface in the kernel scheduling queue; When the timer count reaches the set fixed sampling interval span, the two parallel reading tasks mentioned above are triggered simultaneously. Send a high-level command to the main control pin of the operating device to obtain the real-time analog voltage and current signals on the first return loop; Synchronously send register read commands to the data pins of the auxiliary equipment to obtain the simulated environmental physical quantity signals on the second return loop; The first and second return loop signals are appended with the same nanosecond-level system timestamp to ensure the alignment of the underlying acquisition signals in the time domain.

[0013] Furthermore, the steps of encapsulating the extracted data and verification code into a standard message in Extensible Markup Language (XML) format and pushing it to the production management system for archiving records include: In accordance with the extensible markup language standard specification, create a root node and name it process record; Sequentially generate a time sub-node, an identity sub-node, an operation text command sub-node and a multi-dimensional data vector set sub-node under the root node; Write each extracted corresponding value into the text interval of each newly created sub-node respectively; Read the hash value of the tamper-proof checksum of the previous historical process record, and insert it into the currently newly created root node as an independent traceability field; Attach the currently calculated tamper-proof checksum as a digital signature tag to the end of the root node, and construct an encryption chain including time-dependent relationships; Call the transmission control protocol socket sending interface of the system, serialize the constructed standard message into a byte stream, and push it to the designated network port of the production management system to complete archiving.

[0014] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: Through real-time comparison between voice commands and process requirements, the problem that incorrect installation and missing installation are difficult to detect is solved. After an operator issues a voice command, authority verification and command parsing are performed first, and then comparison with the equipment status constraints of the current process step is carried out, and execution is only allowed when the matching is successful. This process binds process execution to commands in real time, avoids the situation that missing installation and incorrect installation caused by unidentifiable operation behaviors are only exposed after completion, and greatly reduces rework costs.

[0015] On this basis, through continuous verification of equipment status, the problem of disconnection between process execution and equipment status is eliminated. Before an operation command is issued, real-time equipment parameters are automatically compared with process requirements, and a linked operation is triggered only when the equipment status meets the standards. This ensures that the execution of each process is based on compliant equipment, and effectively blocks hidden process violations before they occur.

[0016] At the same time, equipment parameters, operation commands and personnel identities are packaged as process execution logs and uploaded for management, which solves the problem that it is difficult to locate the root cause after an abnormality occurs. Management personnel can trace back to the specific process step, operator and equipment status, which significantly shortens the problem troubleshooting cycle, and forms a closed management loop from error prevention at the source to process control and then to post-event traceability. Description of Drawings

[0017] Figure 1 is a flow chart of the intelligent management method for an industrial Hongmeng workbench based on voice control provided by an embodiment of the present application. Detailed Description of Embodiments

[0018] This application provides a voice-controlled industrial HarmonyOS workbench intelligent management method, which solves the problems in the prior art such as the disconnect between process execution and equipment status, low human-machine interaction efficiency leading to production cycle interruption, and difficulty in tracing specific work steps and operators after anomalies occur. By comparing voice commands with process requirements in real time, continuously verifying equipment status, and automatically generating full-process logs, it realizes intelligent production management from source error prevention and closed-loop process control to accurate post-event traceability.

[0019] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0020] like Figure 1 The diagram shown is a flowchart of an intelligent management method for an industrial HarmonyOS workbench based on voice control provided in an embodiment of this application. The method includes the following steps: collecting operator voice audio, confirming operation permissions and extracting operator identity identifiers through voiceprint comparison, and parsing the voice audio into device operation text commands after the permissions are granted. Establish a data communication link with the production management system, download process operation guidance data, extract the equipment status constraints of the current process step, compare the equipment operation text instructions with the equipment status constraints, and generate a collaborative control instruction set. The collaborative control instruction set is sent to the communication interface of the industrial HarmonyOS controller to trigger the linkage operation of multiple target workbench devices, and the real-time operating parameters of the target workbench devices are synchronously obtained through the communication interface. The system matches real-time operating parameters with equipment status constraints. If the match fails, a voice alarm command is triggered. If the match is successful, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system.

[0021] Furthermore, the steps of confirming operation permissions and extracting operator identification through voiceprint comparison, and then parsing the voice audio into device operation text commands after permission is granted, include: Raw speech audio, including background noise, was captured using an industrial microphone; Silence detection and frame segmentation are performed on the original speech audio, the acoustic feature vectors of each speech frame are extracted, and the first feature matrix is ​​constructed by concatenating them. Read the authorized personnel voiceprint reference matrix library from the system memory, and calculate the similarity distance between the first feature matrix and each matrix in the voiceprint reference matrix library; When the minimum similarity distance is less than the safety threshold, the operation permission is approved, and the corresponding operator's identity is extracted. With permissions granted, the raw voice audio input to the acoustic decoding module is mapped to a Chinese string; Perform word segmentation and part-of-speech tagging on Chinese strings, remove stop words and retain action verbs and target nouns, convert the retained action verbs and target nouns into equipment operation text instructions, and synchronously load the equipment operation text instructions and operator identification into the internal data bus for later use.

[0022] Furthermore, the step of comparing equipment operation text commands with equipment status constraints to generate a collaborative control command set includes: Download the process operation instruction data file for the current product batch; Parse the process operation instruction data file and construct a directed acyclic graph containing multiple process step nodes; Get the current execution timestamp, locate the corresponding current step node in the directed acyclic graph, and read the device type and working parameter constraint range associated with the node; Match the target nouns in the device operation text instructions in the data bus with the device type; When the characters match, the action verb is converted into a low-level control primitive, and supplementary control primitives for linking external monitoring equipment are generated based on the working parameter constraint range. Align and merge the timelines of the underlying control primitives and the supplementary control primitives to generate a collaborative control instruction set.

[0023] Furthermore, the real-time operating parameters of the target workbench device are synchronously acquired through the communication interface: Disassemble the collaborative control instruction set and separate the master control communication data packet and the linkage communication data packet; Through kernel mechanisms, master control communication data packets are routed to the general asynchronous transceiver interface and sent to the master operating device to execute physical actions; Simultaneously, the linkage communication data packets are routed to the local area network bus interface and sent to the auxiliary equipment to initiate the response; during the linkage operation cycle, the data polling process is started; Configure a fixed sampling interval and read the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface; The analog voltage signal is converted into a digital quantized value, and the digital quantized values ​​at the same sampling time are concatenated into a multi-dimensional data vector, which is then stored in the cache as a real-time operating parameter.

[0024] Furthermore, if the matching fails, a voice alarm command is triggered; if the matching succeeds, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system. This process includes: Extract the multidimensional data vector from the cache and use a sliding time window to extract continuous parameter subsequences; Compare the mean of the parameter subsequence with the device state constraint interval; If the mean exceeds the constraint range, a hardware interrupt signal is generated, driving the speaker module to play a voice alarm audio containing the name of the out-of-bounds device; If the mean is within the constraint range, then extract the timestamp, multidimensional data vector, operation text instructions, and operator identification. The tamper-proof verification code of the extracted data is calculated using a hash algorithm; The extracted data and verification code are encapsulated into a standard message in Extensible Markup Language format and pushed to the production management system for archiving records.

[0025] Furthermore, the steps of performing silence detection and frame segmentation on the original speech audio, extracting the acoustic feature vectors of each speech frame, and concatenating them to construct the first feature matrix include: Based on preset window length and window shift parameters, a Hamming window function is applied to the original speech audio to perform frame segmentation, suppress spectral leakage, and obtain multiple overlapping audio short frames. For each audio frame, a fast Fourier transform algorithm is performed to convert the time-domain signal into frequency-domain energy distribution spectrum data. The frequency domain energy distribution spectrum data is passed through a Mel scale filter bank containing multiple triangular filters to calculate the absolute value of the logarithmic energy in each frequency band. Perform a discrete cosine transform on the resulting logarithmic energy absolute value sequence to eliminate the correlation within the data matrix and retain the quantized data of the preceding low-frequency dimensions; All short frames of low-frequency dimensional quantized data are sequentially stitched together in a row-level fashion to generate a two-dimensional numerical array containing the dynamic evolution of acoustics. This array is defined as the first feature matrix for underlying alignment.

[0026] In this embodiment, a significant amount of background noise from mechanical operation exists in a real industrial workshop environment. When the operator issues a voice command, the industrial HarmonyOS controller uses its externally mounted high-noise-resistant industrial array microphone to collect ambient audio in real time and cache it in a memory buffer.

[0027] To accurately separate audio frames containing valid instructions and verify operator identity, the controller kernel first calls the audio processing process to sample and quantize the raw analog audio stream, converting it into a digital audio sequence. Then, it performs a framing operation, dividing the continuous audio sequence into several short frames of fixed length (e.g., a frame length set to 20 milliseconds).

[0028] For the extracted audio frames, their acoustic feature vectors need to be calculated, and a first feature matrix for underlying comparison needs to be constructed. To determine whether the current voiceprint belongs to an authorized operator, a pre-recorded authorized personnel voiceprint baseline matrix library is read from a secure storage area (such as TrustZone). Here, an evolutionary model of Euclidean distance is used to calculate the similarity difference value between the current first feature matrix and the baseline matrix. The formula for calculating the similarity difference value is as follows: ; In the formula, This represents the spatial distance difference between the currently extracted feature matrix and the baseline matrix in the comparison database. The smaller the value, the more similar the voiceprints. This represents the total number of dimensions of the acoustic features defined (e.g., the number of dimensions of the extracted Mel frequency cepstral coefficients). The voice audio representing the current operator is in the [number]th [section]. The feature component values ​​in each dimension Represents an authorized person in the voiceprint reference matrix library at the [number]th [location]. The standard feature component values ​​in each dimension.

[0029] The calculated Compare with the set safety threshold. Only when... Only when the value is below the safety threshold is the permission granted, and the corresponding operator identity identifier (such as employee ID) is extracted based on the matched baseline matrix.

[0030] Subsequently, the internal natural language processing engine is invoked to extract key action verbs (such as "start" and "stop") and target nouns (such as "soldering iron" and "exhaust fan") from the speech audio, combine and splice them into computer-readable device operation text instructions, and store them in the internal data bus for the next call.

[0031] Furthermore, the steps for parsing the process instruction data file and constructing a directed acyclic graph containing multiple process step nodes include: Extract the Extensible Markup Language (XML) document structure from the process instruction data file; Traverse the process level tags in the document structure and extract the unique identifier, the set of prerequisite dependency identifiers, and the process parameter thresholds for each operation. Initialize a blank graph data structure in computer memory; Instantiate the unique identifier of each operation as an independent graph node in the graph data structure; Read the set of node prerequisite dependency identifiers and generate connection edges with directional attributes between source graph nodes and target graph nodes based on the dependency relationships. Attach the device type and working parameter constraint range attribute fields to each independent graph node; Perform a depth-first traversal algorithm on the generated graph data structure to verify acyclic logic; If no closed-loop path is detected, a directed acyclic graph containing multiple work step nodes is output to the system cache.

[0032] In this embodiment, after the operator's authentication is successful and an operation text instruction is generated, it is necessary to determine whether the instruction meets the requirements of the current production process. To this end, the controller establishes a data communication link with the upper-level production management (MES) through its Ethernet interface.

[0033] The controller downloads the XML-formatted process instruction data file for the current product batch via HTTP or MQTT protocols. This file contains all the steps in the product assembly, the types of equipment used, and strict sequential dependencies. To enable the computer to understand these complex logical relationships, the memory controller initializes a blank data structure in the heap memory and traverses the XML file, extracting each step node and constructing it into a directed acyclic graph (DAG).

[0034] When constructing a directed acyclic graph (DAG), each process step is instantiated as an independent graph node, and connection edges with directional attributes are generated between the source and target nodes according to the prerequisites in the process document. To prevent logical infinite loops caused by errors in the process document, a depth-first search (DFS) algorithm is performed on the generated graph data structure. The graph is only loaded and made effective after confirming that there are no closed-loop paths in it.

[0035] Subsequently, the current step node in the directed acyclic graph is located, and the constraint range of the working parameters attached to that node is read (e.g., specifying the lower and upper limits of the soldering temperature). The device operation text instructions in the data bus are compared with the device type required by that node. After a match, to ensure that multiple device actions take effect at the same time, a time-axis alignment operation is performed on the generated low-level control primitives and supplementary control primitives to generate a cooperative control instruction set. The calculation logic for time-axis alignment is as follows: ; In the formula, This indicates the target timestamp for the unified startup or execution of all associated devices within the collaborative control command set. This indicates the base timestamp indicating when the industrial HarmonyOS controller has received and parsed a valid operation command. This indicates the estimated time taken for data processing on the controller's internal bus (which can be estimated by reading the processor clock frequency). This indicates the compensation delay duration set for different peripheral devices (used to smooth out differences in hardware response speeds between different devices).

[0036] Furthermore, the steps of configuring a fixed sampling interval and reading the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface include: Read the current count value of the high-precision hardware timer in the industrial HarmonyOS controller kernel and set the polling reference time based on microseconds; Register parallel read tasks for the general asynchronous transceiver interface and the local area network bus interface in the kernel scheduling queue; When the timer count reaches the set fixed sampling interval span, the two parallel reading tasks mentioned above are triggered simultaneously. Send a high-level command to the main control pin of the operating device to obtain the real-time analog voltage and current signals on the first return loop; Synchronously send register read commands to the data pins of the auxiliary equipment to obtain the simulated environmental physical quantity signals on the second return loop; The first and second return loop signals are appended with the same nanosecond-level system timestamp to ensure the alignment of the underlying acquisition signals in the time domain.

[0037] In this embodiment, after generating the collaborative control instruction set, the industrial HarmonyOS controller uses the kernel's routing mechanism to send the main control communication data packet and the linkage communication data packet to the general asynchronous transceiver interface (such as the RS485 bus interface) and the local area network bus interface, respectively, to trigger the physical actions of devices such as soldering irons and electrostatic wrist strap monitors.

[0038] To monitor equipment compliance in real time, a high-frequency data polling process is initiated simultaneously with the issuance of commands. The controller's microsecond-level hardware timer is read, and a strictly fixed sampling interval is configured. When the timer triggers an interrupt, the controller synchronously sends level reading commands to the sensor pins of both the operating and auxiliary devices to obtain continuous analog voltage signals returned by the sensors.

[0039] Since the sensor returns analog signals, they must be converted into digital quantized values ​​via a built-in analog-to-digital converter (ADC) and further restored to real-time operating parameters with physical meaning. The formula for converting analog signals to real physical quantities in engineering applications is as follows: ; In the formula, This indicates the actual physical parameters of the restored equipment. This represents the discrete digital quantized value read from the analog-to-digital conversion interface (e.g., the integer value read from the ADC register). This indicates the resolution bit depth of the analog-to-digital conversion interface (e.g., for a 12-bit ADC). ), This indicates the reference voltage (in volts) connected when the analog-to-digital converter module is operating. ), This represents the characteristic conversion coefficient of the corresponding hardware sensor (e.g., the coefficient of a thermocouple sensor). Multiple values ​​obtained from the same sampling time are converted... The data is concatenated into a multidimensional data vector and stored in a cache.

[0040] Furthermore, the steps of encapsulating the extracted data and verification code into a standard message in Extensible Markup Language (XML) format and pushing it to the production management system for archiving records include: Create a root node and name it "Process Record" according to the Extensible Markup Language standard specification; Under the root node, generate time sub-nodes, identity sub-nodes, operation text instruction sub-nodes, and multidimensional data vector set sub-nodes in sequence; Write the extracted values ​​into the text range of each newly created child node; Read the tamper-proof verification code hash value from the previous historical process record and insert it as an independent traceability field into the newly created root node; The currently calculated tamper-proof verification code is appended as a digital signature tag to the end of the root node to construct an encrypted chain containing time dependencies. The system's Transmission Control Protocol (TCP) socket sending interface is invoked to serialize the constructed standard message into a byte stream and push it to the designated network port of the production management system for archiving.

[0041] In this embodiment, after obtaining the multidimensional data vector, the values ​​from a single sample cannot be directly compared with the device state constraints because electromagnetic interference in industrial environments can easily cause sudden changes in single sensor data. Therefore, the multidimensional data vector in the cache is extracted, and a sliding time window algorithm is used to extract continuous parameter subsequences, and the smoothed average value within this time window is calculated. The sliding window smoothing formula is as follows: ; In the formula, This represents the current window's integrated device parameter values ​​after anti-interference smoothing processing. This indicates the length of the set sliding time window (i.e., the number of sampling points included, such as 10 points collected continuously, dimensionless). This indicates the first consecutive historical step backwards within a time window. The specific values ​​of a multidimensional data vector.

[0042] Calculated A rigorous comparison is made with the device state constraints (upper and lower limits of the interval) extracted earlier.

[0043] like If the equipment falls within the constraint range, it proves that the current operating status of the equipment fully meets the process requirements and there is no hidden violation.

[0044] Subsequently, the log encapsulation and anti-tampering processing flow begins. To ensure that the root cause can be located quickly and reliably after an anomaly occurs, the current timestamp, multidimensional data vector, device operation text commands, and operator identification are extracted and merged, and a hash algorithm is used to calculate the anti-tampering verification code.

[0045] When calculating the current log checksum, it is mandatory to include the checksum recorded in the previous historical operation as an input variable in the calculation. The calculation formula is as follows: ; In the formula, A unique digital digest label representing the process execution log currently being generated. This indicates the invocation of the secure hash calculation function. This represents the byte stream collection of currently extracted real-time parameters and operator data. This indicates a sequential concatenation operation performed on two byte streams in computer memory. This represents the digital summary label of the previous completed process log read from local storage.

[0046] This nested checksum calculation method constructs an encrypted chain at the underlying data logic level. Finally, the extracted data and checksum are encapsulated into a standard message in Extensible Markup Language (XML) format, serialized into a byte stream via the TCP Transmission Control Protocol (TCP) socket interface, and pushed to the production management archive. Since any modification to a parameter of a historical process will cause the hash value verification of all subsequent processes to fail, this embodiment fundamentally ensures the high reliability and immutability of production traceability data.

[0047] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0049] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0052] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A voice-controlled intelligent management method for industrial HarmonyOS workbenches, characterized in that: Includes the following steps: Collect operator voice audio, confirm operation permissions and extract operator identity identifier through voiceprint comparison, and parse the voice audio into device operation text instructions after the permissions are granted; Establish a data communication link with the production management system, download process operation guidance data, extract the equipment status constraints of the current process step, compare the equipment operation text instructions with the equipment status constraints, and generate a collaborative control instruction set. The steps for generating a cooperative control instruction set include: Download the process operation instruction data file for the current product batch; Parse the work instruction data file and construct a directed acyclic graph containing multiple work step nodes; Obtain the current execution timestamp, locate the corresponding current step node in the directed acyclic graph, and read the device type and working parameter constraint range associated with the node; Match the target nouns in the device operation text instructions in the data bus with the device type; When the characters match, the action verb is converted into a low-level control primitive, and supplementary control primitives for linking external monitoring equipment are generated based on the working parameter constraint range. Align and merge the timelines of the underlying control primitives and the supplementary control primitives to generate a collaborative control instruction set. The collaborative control instruction set is sent to the communication interface of the industrial HarmonyOS controller to trigger the linkage operation of multiple target workbench devices, and the real-time operating parameters of the target workbench devices are synchronously obtained through the communication interface. The system matches real-time operating parameters with equipment status constraints. If the match fails, a voice alarm command is triggered. If the match is successful, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system.

2. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 1, characterized in that, The steps of confirming operation permissions and extracting operator identification through voiceprint comparison, and then parsing the voice audio into device operation text commands after permission is granted, include: Raw speech audio, including background noise, was captured using an industrial microphone; Silence detection and frame segmentation are performed on the original speech audio, the acoustic feature vectors of each speech frame are extracted, and the first feature matrix is ​​constructed by concatenating them. Read the authorized personnel voiceprint reference matrix library from the system memory, and calculate the similarity distance between the first feature matrix and each matrix in the voiceprint reference matrix library; When the minimum similarity distance is less than the safety threshold, the operation permission is approved, and the corresponding operator's identity is extracted. With permissions granted, the raw voice audio input to the acoustic decoding module is mapped to a Chinese string; Perform word segmentation and part-of-speech tagging on Chinese strings, remove stop words and retain action verbs and target nouns, convert the retained action verbs and target nouns into equipment operation text instructions, and synchronously load the equipment operation text instructions and operator identification into the internal data bus for later use.

3. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 1, characterized in that, Real-time operating parameters of the target workbench device are obtained synchronously via the communication interface. Disassemble the collaborative control instruction set and separate the master control communication data packet and the linkage communication data packet; Through kernel mechanisms, master control communication data packets are routed to the general asynchronous transceiver interface and sent to the master operating device to execute physical actions; Simultaneously, the linkage communication data packets are routed to the local area network bus interface and sent to the auxiliary equipment to initiate the response; during the linkage operation cycle, the data polling process is started; Configure a fixed sampling interval and read the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface; The analog voltage signal is converted into a digital quantized value, and the digital quantized values ​​at the same sampling time are concatenated into a multi-dimensional data vector, which is then stored in the cache as a real-time operating parameter.

4. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 1, characterized in that, If the match fails, a voice alarm command is triggered. If the match succeeds, the real-time operating parameters, equipment operation text commands, and operator identification are packaged into a process execution log and uploaded to the production management system. The steps include: Extract the multidimensional data vector from the cache and use a sliding time window to extract continuous parameter subsequences; Compare the mean of the parameter subsequence with the device state constraint interval; If the mean exceeds the constraint range, a hardware interrupt signal is generated, driving the speaker module to play a voice alarm audio containing the name of the out-of-bounds device; If the mean is within the constraint range, then extract the timestamp, multidimensional data vector, operation text instructions, and operator identification. The tamper-proof verification code of the extracted data is calculated using a hash algorithm; The extracted data and verification code are encapsulated into a standard message in Extensible Markup Language format and pushed to the production management system for archiving records.

5. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 2, characterized in that, The steps of performing silence detection and frame segmentation on the original speech audio, extracting the acoustic feature vectors of each speech frame, and concatenating them to construct the first feature matrix include: Based on preset window length and window shift parameters, a Hamming window function is applied to the original speech audio to perform frame segmentation, suppress spectral leakage, and obtain multiple overlapping audio short frames. For each audio frame, a fast Fourier transform algorithm is performed to convert the time-domain signal into frequency-domain energy distribution spectrum data. The frequency domain energy distribution spectrum data is passed through a Mel scale filter bank containing multiple triangular filters to calculate the absolute value of the logarithmic energy in each frequency band. Perform a discrete cosine transform on the resulting logarithmic energy absolute value sequence to eliminate the correlation within the data matrix and retain the quantized data of the preceding low-frequency dimensions; All short frames of low-frequency dimensional quantized data are sequentially stitched together in a row-level fashion to generate a two-dimensional numerical array containing the dynamic evolution of acoustics. This array is defined as the first feature matrix for underlying alignment.

6. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 1, characterized in that, The steps for parsing the work instruction data file and constructing a directed acyclic graph containing multiple work step nodes include: Extract the Extensible Markup Language (XML) document structure from the process instruction data file; Traverse the process level tags in the document structure and extract the unique identifier, the set of prerequisite dependency identifiers, and the process parameter thresholds for each operation. Initialize a blank graph data structure in computer memory; Instantiate the unique identifier of each operation as an independent graph node in the graph data structure; Read the set of node prerequisite dependency identifiers and generate connection edges with directional attributes between source graph nodes and target graph nodes based on the dependency relationships. Attach the device type and working parameter constraint range attribute fields to each independent graph node; Perform a depth-first traversal algorithm on the generated graph data structure to verify acyclic logic; If no closed-loop path is detected, a directed acyclic graph containing multiple work step nodes is output to the system cache.

7. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 3, characterized in that, The steps for configuring a fixed sampling interval and reading the analog voltage signals returned by the sensors of the operating equipment and auxiliary equipment through the analog-to-digital conversion interface include: Read the current count value of the high-precision hardware timer in the industrial HarmonyOS controller kernel and set the polling reference time based on microseconds; Register parallel read tasks for the general asynchronous transceiver interface and the local area network bus interface in the kernel scheduling queue; When the timer count reaches the set fixed sampling interval span, the two parallel reading tasks mentioned above are triggered simultaneously. Send a high-level command to the main control pin of the operating device to obtain the real-time analog voltage and current signals on the first return loop; Synchronously send register read commands to the data pins of the auxiliary equipment to obtain the simulated environmental physical quantity signals on the second return loop; The first and second return loop signals are appended with the same nanosecond-level system timestamp to ensure the alignment of the underlying acquisition signals in the time domain.

8. The intelligent management method for an industrial HarmonyOS workbench based on voice control as described in claim 4, characterized in that, The steps involved in encapsulating the extracted data and verification code into a standard message in Extensible Markup Language (XML) format and pushing it to the production management system for archiving records include: Create a root node and name it "Process Record" according to the Extensible Markup Language standard specification; Under the root node, generate time sub-nodes, identity sub-nodes, operation text instruction sub-nodes, and multidimensional data vector set sub-nodes in sequence; Write the extracted values ​​into the text range of each newly created child node; Read the tamper-proof verification code hash value from the previous historical process record and insert it as an independent traceability field into the newly created root node; The currently calculated tamper-proof verification code is appended as a digital signature tag to the end of the root node to construct an encrypted chain containing time dependencies; The system's Transmission Control Protocol (TCP) socket sending interface is invoked to serialize the constructed standard message into a byte stream and push it to the designated network port of the production management system for archiving.

Images