An industrial equipment control method and electronic device based on multi-source heterogeneous data

Abstract

The embodiment of the application discloses an industrial equipment control method and electronic equipment based on multi-source heterogeneous data, relates to the technical field of intelligent control, and the method comprises the following steps: performing cross-modal feature fusion processing on the working condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain a fusion feature matrix of the industrial equipment; performing working condition state observation calculation on the fusion feature matrix to obtain an estimated working condition data sequence of the industrial equipment; determining a control strategy of the industrial equipment according to the working condition data sequence and the estimated working condition data sequence, and controlling the industrial equipment based on the control strategy. Through the technical scheme of the embodiment of the application, more accurate, reliable and intelligent control of the industrial equipment can be realized.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of intelligent control technology, and in particular to an industrial equipment control method and electronic device based on multi-source heterogeneous data. Background Technology

[0002] With the rapid development of new-generation information technologies such as the Industrial Internet, big data, and artificial intelligence, industrial manufacturing is undergoing a profound evolution towards digitalization, networking, and intelligence. As the core carrier of the production system, the effective perception and precise control of the operating status of industrial equipment directly affects production efficiency, product quality, energy consumption, and operational safety.

[0003] Currently, industrial equipment control methods primarily rely on real-time monitoring of operating condition data such as voltage, current, and temperature, and make control responses based on fixed thresholds. For example, when the equipment's operating temperature exceeds a set threshold, the cooling system is directly activated or an alarm is triggered. However, this type of method, depending solely on operating condition data, makes it difficult to comprehensively assess the overall operating status of the equipment. This can easily lead to one-sided control decisions and insufficient control precision, failing to meet the refined and intelligent control requirements of industrial equipment.

[0004] Therefore, there is an urgent need to propose a new method to solve the above problems. Summary of the Invention

[0005] This invention provides an industrial equipment control method and electronic device based on multi-source heterogeneous data, which can achieve more precise, reliable and intelligent control of industrial equipment.

[0006] In a first aspect, embodiments of the present invention provide an industrial equipment control method based on multi-source heterogeneous data, the method comprising: Cross-modal feature fusion processing is performed on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment. The fused feature matrix is ​​subjected to operating condition observation and calculation to obtain the estimated operating condition data sequence of the industrial equipment. The control strategy for the industrial equipment is determined based on the operating condition data sequence and the estimated operating condition data sequence, and the industrial equipment is controlled based on the control strategy.

[0007] The technical solution of this invention first performs cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of industrial equipment within the current period to obtain a fused feature matrix of the industrial equipment. This breaks the limitations of single, homogeneous data representation and takes into account both quantitative operating condition parameters and textual operation record information of the industrial equipment, thus comprehensively restoring the true operating state of the industrial equipment. Based on this, the overall operating state of the equipment can be comprehensively evaluated, effectively improving the completeness and accuracy of equipment state representation, while enhancing the sensitivity of industrial equipment operation monitoring. This provides a standardized and reliable data foundation for subsequently determining the estimated operating condition data sequence of the industrial equipment. Operating condition observation and calculation are then performed on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment. This outputs an accurate and continuous estimated operating condition data sequence, providing reliable time-series data support for subsequent industrial equipment control strategy formulation, thereby improving the accuracy and reliability of equipment control and ensuring the safe and stable operation of the industrial equipment. This invention determines the control strategy for industrial equipment based on operating condition data sequences and estimated operating condition data sequences, and controls the equipment based on these strategies. This avoids the risk of misjudgment that arises from relying solely on measured operating condition data for decision-making and control, thus improving the accuracy and reliability of control strategy formulation. It achieves fully automated closed-loop control of the industrial equipment's operating status, reducing manual intervention and improving the timeliness and intelligence of equipment control. This ensures the stable and efficient operation of the industrial equipment and enhances the overall precision and practicality of the control. Therefore, this invention overcomes the shortcomings of existing technologies that rely solely on measured operating condition data, making it difficult to comprehensively assess the overall operating status of the equipment. This can lead to one-sided and inaccurate control decisions, failing to meet the needs of refined and intelligent control of industrial equipment.

[0008] Secondly, embodiments of the present invention also provide an industrial equipment control device based on multi-source heterogeneous data, the device comprising: The fusion module is used to perform cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment. The calculation module is used to perform operating condition observation and calculation on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment; The control module is used to determine the control strategy of the industrial equipment based on the operating condition data sequence and the estimated operating condition data sequence, and to control the industrial equipment based on the control strategy.

[0009] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform the industrial equipment control method based on multi-source heterogeneous data according to any embodiment of the present invention.

[0010] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, characterized in that the computer-executable instructions, when executed by a computer processor, implement the industrial equipment control method based on multi-source heterogeneous data as described in any embodiment of the present invention.

[0011] It should be noted that the aforementioned computer instructions may be stored, in whole or in part, on a computer-readable storage medium. This computer-readable storage medium may be packaged together with the processor of the industrial equipment control device based on multi-source heterogeneous data, or it may be packaged separately from the processor of the industrial equipment control device based on multi-source heterogeneous data; this application does not impose any limitations on this.

[0012] The descriptions of the second, third, and fourth aspects in this application can be referenced to the detailed description of the first aspect; and the beneficial effects described in the second, third, and fourth aspects can be referenced to the analysis of the beneficial effects of the first aspect, which will not be repeated here.

[0013] In this application, the names of the aforementioned industrial equipment control devices based on multi-source heterogeneous data do not limit the devices or functional modules themselves. In actual implementation, these devices or functional modules may appear under other names. As long as the functions of each device or functional module are similar to those in this application, they fall within the scope of the claims of this application and their equivalents.

[0014] These or other aspects of this application will become more readily apparent in the following description. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating an industrial equipment control method based on multi-source heterogeneous data, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another industrial equipment control method based on multi-source heterogeneous data provided in an embodiment of the present invention; Figure 3A schematic diagram of the structure of an industrial equipment control device based on multi-source heterogeneous data provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0018] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0019] The terms "first" and "second," etc., used in the specification and drawings of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.

[0020] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0021] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc. Moreover, without conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0022] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0023] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0024] Figure 1 This is a flowchart illustrating an industrial equipment control method based on multi-source heterogeneous data, provided by an embodiment of the present invention. This embodiment is applicable to scenarios involving intelligent automated control of industrial equipment. The method is executed by an industrial equipment control device based on multi-source heterogeneous data, which can be implemented in software and / or hardware. For example, the device can be integrated into an electronic device. The industrial equipment control method based on multi-source heterogeneous data in this embodiment specifically includes the following steps: Step 110: Perform cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of industrial equipment in the current period to obtain the fusion feature matrix of industrial equipment.

[0025] Specifically, industrial equipment refers to a collective term for various machines, devices, units, and supporting sensing and control devices used in industrial production, processing, operation, maintenance, and monitoring scenarios, possessing functions such as mechanical operation, energy conversion, material handling, operating condition sensing, and automatic control. For example, industrial equipment can refer to electrical equipment in a power system. A power system is a unified whole composed of power plants, transmission networks, distribution networks, and power users, used to complete the production, transmission, distribution, and consumption of electrical energy. Electrical equipment refers to various devices used in a power system for power generation, transmission, transformation, distribution, and supply, such as energy storage batteries, inverters, photovoltaic modules, transformers, circuit breakers, switchgear, generator sets, reactors, capacitors, and transmission line equipment. The current period is a fixed time window or sampling interval pre-set according to the equipment's operating requirements. For example, a period can be 1 minute, 1 hour, or 1 day, used to uniformly limit the scope of operating condition data and operational text data extraction, ensuring data timing consistency. Operating condition data refers to structured numerical monitoring data collected during the operation of industrial equipment, such as time-series quantified parameters like temperature, current, voltage, load, vibration amplitude, insulation resistance, oil temperature, and power factor. The operating condition data sequence is the raw data set obtained by arranging the operating condition data collected within the current period in chronological order. Equipment operation text data refers to unstructured text data describing the operating status of industrial equipment, such as equipment alarm logs, maintenance and inspection records, fault description text, operation notes, and abnormal event text records. Cross-modal feature fusion is the process of extracting effective features from multi-source heterogeneous raw information from different data modalities (such as numerical operating condition data and text-based operation text data) and integrating them into a unified and complementary feature representation through algorithms. The fused feature matrix is ​​a two-dimensional array data structure generated after cross-modal feature fusion, carrying comprehensive equipment operation information; for example, rows of the matrix can correspond to different time points or samples, and columns correspond to the fused multi-dimensional features.

[0026] In the specific implementation, firstly, based on a pre-set data acquisition cycle, two types of raw data corresponding to the industrial equipment within the current cycle are synchronously captured: operating condition data and equipment operation text data. For the operating condition data, it is first preprocessed (e.g., outlier removal, missing value imputation, and standardization). Then, the preprocessed operating condition data is organized and sorted according to the time dimension to form an operating condition data sequence for the industrial equipment. Next, a time-series statistical feature extraction algorithm (e.g., frequency domain statistical feature extraction algorithm) is used to extract features from the operating condition data sequence to obtain operating condition features. For the equipment operation text data, it is first preprocessed (e.g., text deduplication, invalid character filtering, semantic segmentation, and stop word removal). Then, a text semantic extraction model (e.g., BERT semantic extraction model) is used to extract features from the preprocessed operation text data to obtain operation text features. Finally, a cross-modal fusion algorithm (e.g., cross-attention fusion algorithm, multi-head attention cross-modal fusion algorithm, self-attention cross-modal fusion algorithm, etc.) is used to fuse the operating condition features and operation text features to obtain the fused feature matrix of the industrial equipment.

[0027] It should be noted that in practical applications, equipment operation text data is usually event-triggered sparse data, and not every time step (a time unit with a fixed time interval formed after discretizing the operating condition data sequence) has a corresponding text record. Therefore, when performing cross-modal fusion of operating condition features and operation text features, text features are only incorporated into the corresponding time steps with text records and their adjacent time steps; for the remaining time steps without text records, the text features are set to zero vectors before participating in feature fusion.

[0028] In this embodiment, cross-modal feature fusion processing is performed on multi-source heterogeneous data (operating condition data sequences and equipment operation text data) of industrial equipment to obtain a fused feature matrix of industrial equipment. This breaks the limitations of single homogeneous data representation and takes into account both quantitative operating condition parameters and text-based operation record information of industrial equipment, thus comprehensively restoring the true operating status of industrial equipment. On this basis, the overall operating status of the equipment can be comprehensively evaluated, effectively improving the completeness and accuracy of equipment status representation, while enhancing the sensitivity of industrial equipment operation monitoring, and providing a standardized and reliable data foundation for subsequently determining the estimated operating condition data sequence of industrial equipment.

[0029] Step 120: Perform operating condition observation and calculation on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment.

[0030] Specifically, the operational condition observation and calculation refers to the process of using a fused feature matrix as input, employing filtering algorithms such as state observers and Kalman filters, or deep learning models such as Transformers, to calculate and back-derive the actual operating condition variation law of industrial equipment within the current cycle, thereby obtaining the corresponding operational condition sequence. The estimated operational condition data sequence is the time-series predicted operational condition data for the current cycle derived from the operational condition observation and calculation. It is arranged continuously in chronological order and can be compared and referenced with the actual operational condition data sequence collected by field sensors.

[0031] In practice, the fused feature matrix can be input into a pre-trained estimation model to infer the estimated operating condition data sequence of industrial equipment. The estimation model refers to a model that converges after supervised training of a deep learning model such as a temporal convolutional network, using fused feature matrices corresponding to different historical periods and matching real operating condition data as training samples.

[0032] In this embodiment, the above steps can output accurate and continuous estimated operating condition data sequences, providing reliable time-series data support for the subsequent formulation of control strategies for industrial equipment, thereby improving the accuracy and reliability of equipment control and ensuring the safe and stable operation of industrial equipment.

[0033] Step 130: Determine the control strategy for the industrial equipment based on the operating condition data sequence and the estimated operating condition data sequence, and control the industrial equipment based on the control strategy.

[0034] Specifically, the control strategy is a scheme for adjusting the operation mode of industrial equipment based on the comparison and analysis of the measured operating condition data sequence and the estimated operating condition data sequence as the optimal estimate. For example, it may include adjusting the tap of the on-load tap changer, switching capacitor banks, modifying protection settings, issuing early warnings or alarms, adjusting load distribution, triggering the cooling system, etc.

[0035] In practice, the operating condition data sequence and the estimated operating condition data sequence are input into a pre-trained operating condition optimization decision model. This model directly infers and outputs a control strategy adapted to the current operating state of the industrial equipment (such as increasing the cooling fan speed by 20%, lowering the transformer tap by one level, and reducing the bus voltage to the standard range). Then, based on this control strategy, standardized control commands are generated and sent to the corresponding actuators of the industrial equipment. The actuators then perform the corresponding control actions, achieving automatic regulation of the industrial equipment's operating state. After regulation is completed, the process can return to step 110 to continuously cycle through the relevant data acquisition and optimization control process for the industrial equipment.

[0036] The operating condition optimization decision model is an intelligent decision analysis model that uses measured and estimated operating condition sequences as joint inputs. During offline training, the model relies on operating condition constraints (such as prohibiting conflicting control actions during the same time period; disallowing automatic control strategies when equipment is in a locked state; and requiring equipment load rates to follow gradual load changes, prohibiting sudden increases or decreases), equipment operating boundary conditions (such as the maximum allowable operating temperature of transformers or switchgear, upper and lower limits of ambient temperature, bus power flow transmission limits, and critical boundaries of regional power grid voltage stability), and optimization algorithms (such as particle swarm optimization, genetic algorithms, simulated annealing algorithms, multi-objective weighted optimization, and analytic hierarchy process). It is trained based on historical operating conditions and corresponding optimal control samples. After training, the model can directly match and generate control schemes adapted to the current equipment operating state during online inference.

[0037] In this embodiment, the above steps avoid the risk of misjudgment caused by relying solely on measured operating condition data for decision-making and control, thereby improving the accuracy and reliability of control strategy formulation; fully automatic closed-loop control of industrial equipment operating status is realized, reducing manual intervention, thereby improving the timeliness and intelligence level of equipment control, ensuring stable and efficient operation of industrial equipment, and enhancing the accuracy and practicality of overall control.

[0038] In addition, the execution entity in this embodiment can be a controller, edge computing terminal, industrial control computer, or field gateway deployed locally on the industrial equipment, or a control server or cloud operation and maintenance management platform deployed in the background or cloud, which can realize simultaneous monitoring and intelligent control of multiple industrial devices.

[0039] The industrial equipment control method based on multi-source heterogeneous data provided in this invention first performs cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain a fused feature matrix of the industrial equipment. This breaks the limitations of single homogeneous data representation and takes into account both quantitative operating condition parameters and textual operation record information of the industrial equipment, thus comprehensively restoring the true operating state of the industrial equipment. On this basis, the overall operating state of the equipment can be comprehensively evaluated, effectively improving the completeness and accuracy of the equipment state representation, while enhancing the sensitivity of industrial equipment operation monitoring, and providing a standardized and reliable data foundation for subsequently determining the estimated operating condition data sequence of the industrial equipment. Operating condition observation and calculation are performed on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment, which can output accurate and continuous estimated operating condition data sequences, providing reliable time-series data support for subsequent industrial equipment control strategy formulation, thereby improving the accuracy and reliability of equipment control and ensuring the safe and stable operation of industrial equipment. This invention determines the control strategy for industrial equipment based on operating condition data sequences and estimated operating condition data sequences, and controls the equipment based on these strategies. This avoids the risk of misjudgment that arises from relying solely on measured operating condition data for decision-making and control, thus improving the accuracy and reliability of control strategy formulation. It achieves fully automated closed-loop control of the industrial equipment's operating status, reducing manual intervention and improving the timeliness and intelligence of equipment control. This ensures the stable and efficient operation of the industrial equipment and enhances the overall precision and practicality of the control. Therefore, this invention overcomes the shortcomings of existing technologies that rely solely on measured operating condition data, making it difficult to comprehensively assess the overall operating status of the equipment. This can lead to one-sided and inaccurate control decisions, failing to meet the needs of refined and intelligent control of industrial equipment.

[0040] Figure 2 This is a flowchart illustrating another industrial equipment control method based on multi-source heterogeneous data provided in this embodiment of the invention. This embodiment is a specific implementation based on the above embodiment. In this embodiment, the method may further include: Step 210: Perform semantic quantization mapping on the equipment operation text data of industrial equipment in the current cycle to obtain text semantic feature vectors.

[0041] Specifically, semantic quantization mapping is the process of converting natural language text into numerical semantic vectors. A text semantic feature vector refers to a numerical vector obtained after semantic quantization mapping of the device's running text data.

[0042] In the specific implementation, the collected device operation text data is first preprocessed: text cleaning, invalid character removal, stop word filtering, and domain semantic segmentation are performed sequentially to obtain a list of short segmented texts for each text. Then, the short segmented text lists for each text are sequentially input into a pre-trained semantic extraction model (such as BERT-base-Chinese) to obtain a text vector for each text. Finally, the text vectors corresponding to all texts in the current period are fused (e.g., average pooling, max pooling, etc.) to obtain the text semantic feature vector.

[0043] In this embodiment, the above steps transform unstructured text data into structured feature representations, which facilitates cross-modal fusion operations with operating condition time series data. At the same time, it overcomes the drawbacks of traditional solutions that rely solely on numerical operating condition data and ignore the implicit semantic information of text, enabling a more comprehensive and complete characterization of the overall operating status of industrial equipment, thereby improving the accuracy of subsequent equipment control decisions.

[0044] Further, step 210 may specifically include: determining semantic items in the device operation text data; determining the fuzzy semantic quantization parameters corresponding to the semantic items based on preset fuzzy semantic mapping rules; generating a preset number of random simulated samples based on the fuzzy semantic quantization parameters to obtain a fuzzy feature simulation dataset of the semantic items; and extracting features from the fuzzy feature simulation dataset to obtain a text semantic feature vector.

[0045] Specifically, a semantic term is the smallest unit of information extracted from the equipment operation text data that can independently express the equipment's operating status, such as a temperature that is too high. Preset fuzzy semantic mapping rules are a rule system for semantic understanding in industrial fields (such as the power industry), pre-defined according to actual conditions or needs. These rules are used to establish the corresponding mapping relationship between text semantic terms and fuzzy semantic quantization parameters, realizing the rule-based transformation from natural language semantics to fuzzy numerical space. Fuzzy semantic quantization parameters are a set of feature parameters obtained through fuzzy mapping rules, used to quantitatively describe the distribution characteristics of the fuzzy set to which a single semantic term belongs. Random simulated samples are discrete simulated numerical sample points generated through random sampling based on the fuzzy semantic quantization parameters. The preset sample quantity is the total number of randomly simulated samples generated in advance. The fuzzy feature simulation dataset is a sample set composed of all random simulated samples generated for the same semantic term according to the preset sample quantity (e.g., 1000).

[0046] In the specific implementation, firstly, the collected industrial equipment operation text data is cleaned, invalid characters are removed, and semantic word segmentation is performed to filter and determine various semantic items that can independently represent the equipment's operating status. Then, for each semantic item, its corresponding fuzzy semantic quantization parameters [i.e., (a, b, c, d; w)] are determined based on preset fuzzy semantic mapping rules, where a and b are the upper and lower bound parameters of the semantic item; c and d are the kernel interval parameters of the semantic item; and w∈[0,1] is the confidence weight, representing the confidence level of the semantic item]. Next, based on the fuzzy semantic quantization parameters and membership function, the confidence distribution of the semantic item in different parameter intervals is determined; and weighted random sampling is performed according to this distribution (more samples are generated in intervals with higher confidence) to generate a preset number of random simulated samples, obtaining a fuzzy feature simulation dataset for the semantic item. Finally, the membership contribution of each simulated sample in the fuzzy feature simulation dataset in all α-cut sets (e.g., step size 0.01, α range 0-1) is calculated to obtain the confidence of each simulated sample. Next, based on each simulated sample and its corresponding confidence level, the centroid coordinates of the semantic term are calculated and expanded into a feature vector of a preset dimension (e.g., 16 dimensions), thus obtaining the semantic feature vector of the semantic term. Finally, the semantic feature vectors of each semantic term are fused to obtain the text semantic feature vector.

[0047] The membership function is: ; is the membership function value of the current simulated sample value; o is the current simulated sample value.

[0048] For example, the structure of the preset fuzzy semantic mapping rule is: Device Type - Monitoring Parameter - Semantic Item - Parameter Domain U - Fuzzy Semantic Quantization Parameter (a, b, c, d; w). Here, the parameter domain U is the range of values ​​for the operating parameters, a and b are the upper and lower bound parameters of the semantic item, c and d are the kernel interval parameters of the semantic item, and w∈[0,1] is the confidence weight. For example: if the device type is an energy storage battery, the monitoring parameter is surface temperature, the semantic item is slightly warm, the parameter domain U is [0, 80], and the corresponding fuzzy semantic quantization parameter is (20, 30, 40, 45; 0.9); if the device type is an energy storage battery, the monitoring parameter is surface temperature, the semantic item is slightly high, the parameter domain U is [0, 80], and the corresponding fuzzy semantic quantization parameter is (40, 45, 55, 60; 0.95).

[0049] Optionally, the construction rules and engineering calibration steps for the preset fuzzy semantic mapping rules are as follows: The parameter construction rules are as follows: 1. Upper and lower bound parameters a and d: the minimum and maximum physical boundary values ​​of the corresponding semantic item. Where a is the minimum value of the parameter described by the semantic item, and d is the maximum value; the values ​​must not exceed the physical range of the monitored parameter. 2. Core interval parameters b and c: the core membership interval of the corresponding semantic item, i.e., the most typical value range of the parameter described by the semantic item. Within the interval, the membership degree is always equal to the confidence weight w, and must conform to the general understanding of on-site maintenance personnel. 3. Confidence weight w: the value range is [0, 1], and the value is positively correlated with the security risk level corresponding to the semantic item. The higher the risk level, the closer w is to 1.0. For regular descriptive semantic items, w is not less than 0.8; for early warning semantic items, w is not less than 0.9; and for alarm semantic items, w is fixed at 1.0.

[0050] The engineering calibration steps are as follows: 1. Determine the physical range and safety boundary of the monitoring parameter, and clarify the value range of U; 2. Collect the corresponding field operation and maintenance logs, equipment manuals, safety procedures and historical fault data for the parameter, and sort them into general semantic items; 3. Organize 3 or more engineers with more than 5 years of relevant equipment operation and maintenance experience to independently calibrate the parameter range corresponding to each semantic item; 4. Perform consistency verification on multiple sets of calibration results, remove abnormal data with a deviation of more than 20%, and take the mean of the remaining data as the initial value of a, b, c, and d; 5. Verify the initial parameters based on historical fault cases and operation and maintenance records. If the matching degree between the fuzzy number mapping result and the actual working condition is less than 90%, repeat steps 3-4 until the matching degree meets the requirements; 6. Determine the final value of the confidence weight w according to the risk level of the semantic item, and complete the construction of the fuzzy semantic mapping rule.

[0051] In this embodiment, the above steps achieve standardized modeling of fuzzy language, transforming unstructured natural language descriptions into computable structured parameters. At the same time, by generating simulated samples that fit the fuzzy semantic confidence distribution, edge noise interference is effectively avoided, improving the robustness and representativeness of text semantic feature vectors and providing a reliable text feature foundation for subsequent cross-modal fusion.

[0052] Step 211: Construct a time series matrix of operating conditions based on the operating condition data sequence of industrial equipment in the current period.

[0053] Specifically, the operating condition time series matrix is ​​a two-dimensional time series data matrix based on the operating condition data sequence of industrial equipment in the current period, and arranged in a regular manner according to the time step dimension and the operating condition parameter dimension. It centrally carries the time series characteristics of the equipment operating parameters changing over time.

[0054] In practice, the operating condition data sequence of industrial equipment in the current cycle is time-series aligned and regularized according to a preset fixed time step. Missing data is filled in, and outliers are removed, thus obtaining a standardized operating condition parameter sequence with equal time intervals and regular temporal arrangement. Then, the standardized operating condition parameter sequence is mapped into a two-dimensional data structure with time step as the row dimension and operating condition parameter type as the column dimension to obtain the operating condition time series matrix.

[0055] In this embodiment, the above steps can fully characterize the temporal evolution of various operating parameters of industrial equipment, improve the completeness and refinement of equipment operating status characterization, and thus improve the accuracy of equipment control decisions.

[0056] Step 212: Use the target attention mechanism to perform cross-modal attention fusion calculation on the text semantic feature vector and the working condition time series matrix to obtain the fusion feature matrix of industrial equipment.

[0057] Specifically, the target attention mechanism is an attention mechanism predetermined according to actual conditions or needs, used to achieve feature fusion between text semantic feature vectors and operating condition time series matrices. For example, the target attention mechanism can adopt a multi-head attention mechanism. Cross-modal attention fusion calculation is a calculation process that uses attention mechanisms to perform feature association interaction and information integration on two types of heteromodal data: the operating condition time series matrix of the numerical modality and the semantic feature vector of the text modality.

[0058] In the specific implementation, firstly, based on a target attention mechanism (such as a multi-head attention mechanism), query, key, and value feature projection transformations are performed on the text semantic feature vector and the operating condition time series matrix respectively to establish the correlation mapping relationship between the two modal features. Next, the attention weight calculation formula is used to solve the cross-modal attention weight between the text semantic feature vector and the operating condition time series matrix. Weight coefficients are automatically assigned according to the tightness of feature correlation, emphasizing the strengthening of effective features highly correlated with the equipment operating status and weakening the interference caused by redundant noise features. For example, the formula for calculating attention weight can be expressed as Attention(Q, K, V) = softmax[QK]. T / (d k ) 1 / 2 ]V, where Q is the query matrix; K is the key matrix; V is the value matrix; d k The dimension of the Key vector corresponding to a single attention head; the superscript T indicates matrix transpose.

[0059] Then, the obtained attention weights are used to weight the two types of heteromodal features, and the weighted bimodal features are spliced, fused and dimensionally compressed to obtain a fused feature matrix.

[0060] In this embodiment, the above steps break down the information barriers between textual semantics and operating condition time series data, integrate equipment textual operation and maintenance information with numerical operating parameters, avoid the problem of one-sided representation of single modality features, and enable the fused feature matrix to more comprehensively and accurately restore the real operating status of industrial equipment, thereby improving the accuracy and intelligence level of industrial equipment control decisions.

[0061] Optionally, the query matrix of the target attention mechanism is the working condition time series matrix, the key matrix of the target attention mechanism is obtained by extending the text semantic feature vector, and the value matrix of the target attention mechanism is obtained by extending the text semantic feature vector.

[0062] In the specific implementation, firstly, the working condition time series matrix is ​​directly used as the query matrix of the target attention mechanism, and the working condition time series features are used as the global query benchmark to retrieve related text semantic information. Secondly, the fixed-dimensional text semantic feature vector is subjected to dimensional expansion and matrix mapping processing: according to the row and column structure of the working condition time series matrix, it is subjected to dimensional padding, dimensional expansion, and tensor extension, transforming the one-dimensional text semantic feature vector into a two-dimensional matrix that matches the dimension of the query matrix, thereby constructing the key matrix of the attention mechanism.

[0063] Simultaneously, using the same dimensionality expansion method as the key matrix, the text semantic feature vectors are synchronously extended and transformed to generate a value matrix whose dimensions and structure are consistent with the key matrix.

[0064] In this embodiment, the robustness of text semantic representation is enhanced and overfitting is avoided through the above steps, while the original dynamic differences of the working condition time series matrix are preserved. At the same time, cross-modal attention calculation is simplified, the real-time performance of the operation is improved, and the decoupling and re-fusion of text features and working condition features are realized, thereby obtaining a fusion feature matrix of better quality.

[0065] Step 213: Perform operating condition observation and calculation on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment.

[0066] Further, step 213 may specifically include: determining the current operating condition of the industrial equipment; matching the current operating condition with a preset operating condition observation parameter library to obtain a set of state estimation parameters for the current operating condition; and substituting the fused feature matrix and the set of state estimation parameters into a preset estimation calculation model to obtain a sequence of estimated operating condition data for the industrial equipment.

[0067] Specifically, the current operating condition of industrial equipment refers to its actual operating state within the current cycle, such as light load, rated load, or overload. The preset operating condition observation parameter library is a set of state estimation parameters for different operating conditions of industrial equipment, pre-constructed according to actual conditions or needs. It includes information such as the parameter types, value ranges, and correlations required for state estimation under each operating condition, providing a standardized reference for operating condition matching. For example, the state estimation parameter set may include a linear gain matrix and a sliding mode gain matrix. The linear gain matrix can be obtained by solving linear matrix inequality constraints, while the sliding mode gain matrix can be calculated based on the noise upper bound and reaching law parameters, and then corrected through simulation and debugging. The current operating condition state estimation parameter set is a set of state estimation parameters adapted to the current operating condition of the industrial equipment, obtained through operating condition matching, including coefficients, weights, and constraints required for the estimation model. The estimation calculation model is a pre-built adaptive calculation model for estimating the state of industrial equipment based on its operating condition data. The model takes the fused feature matrix and the state estimation parameter set as input, and outputs an estimated operating condition data sequence of industrial equipment adapted to the current operating condition through the operating condition adaptation operation rules configured in the state estimation parameter set.

[0068] In practice, the obtained operating condition data can be input into the operating condition determination model to obtain the current operating condition of the industrial equipment. Then, based on this current operating condition, a matching is performed in a preset operating condition observation parameter library to obtain the state estimation parameter set for the current operating condition. For example, the state estimation parameter set includes the system state matrix A, the system input matrix B, the linear gain matrix L, the sliding mode gain matrix K1, etc. Afterwards, the fused feature matrix and the state estimation parameter set are substituted into a preset estimation calculation model to obtain the estimated operating condition data sequence X(t) of the industrial equipment. Here, the operating condition determination model refers to the model obtained after training a deep learning model based on different historical operating condition data and corresponding historical operating condition labels.

[0069] For example, the estimation computation model adopts a sliding mode observer structure, and the expression is as follows: X2(t) = AX(t) + Bu + L[(X1(t) - X(t)] + K1 sgn[X1(t)-X(t)]+f; where u is the control input vector, X1(t) is the measured operating condition data sequence, X2(t) is the rate of change of the estimated operating condition data sequence; f is the system nonlinear term that satisfies the Lipschitz condition, used to describe the nonlinear characteristics of industrial equipment during operation; sgn() is the sign function, t is the time variable, representing any moment in the continuous time domain.

[0070] In this embodiment, the above steps improve the accuracy and robustness of the estimated operating condition data sequence and reduce the online computational complexity, thereby improving the real-time performance of industrial equipment condition estimation and making it adaptable to various complex operating conditions.

[0071] Step 214: Calculate the residual of the operating condition based on the operating condition data sequence and the estimated operating condition data sequence.

[0072] Specifically, the operating condition residual is the difference between the measured operating condition data sequence and the estimated operating condition data sequence.

[0073] In the specific implementation, the residual of the operating condition is equal to the operating condition data sequence minus the estimated operating condition data sequence.

[0074] In this embodiment, the deviation between the state estimation result and the actual operating state of the equipment can be quantitatively characterized through the above steps. This provides a basis for accuracy feedback for subsequent closed-loop control and a reliable quantitative judgment index for abnormal operating conditions and fault warning.

[0075] Step 215: Determine whether the absolute value of the residual under the operating condition is greater than the target residual threshold.

[0076] If it is greater than, proceed to step 216; if it is not greater than, proceed to step 217.

[0077] Specifically, the target residual threshold is a pre-set deviation limit based on the accuracy requirements of industrial equipment operating condition estimation and safe operation standards. It serves as a quantitative decision criterion for subsequent differentiated processing.

[0078] In practice, after obtaining the residual of the operating condition, it can be determined whether the absolute value of the residual is greater than the target residual threshold. If it is greater, it indicates a significant deviation between the measured and estimated operating conditions, suggesting a high probability of abnormal operation or potential faults in the equipment. In this case, the operating state of the industrial equipment can be determined to be abnormal, and an alarm message can be sent to the operator's terminal for timely manual intervention. If the residual is not greater, it indicates that the measured and estimated operating conditions are basically consistent, and the current operating state of the equipment is normal or within the allowable range. In this case, surrogate optimization processing can be performed on the estimated operating condition data sequence to obtain the control strategy for the industrial equipment, thereby maintaining and improving operational efficiency.

[0079] In this embodiment, the above steps can provide reliable quantitative criteria for subsequent differentiated treatment.

[0080] Furthermore, before step 215, the method includes: constructing a residual covariance matrix based on the residuals of the operating conditions; calculating the trace of the residual covariance matrix to obtain the overall volatility characteristic; and calculating the target residual threshold based on the overall volatility characteristic and the risk critical coefficient.

[0081] Specifically, the residual covariance matrix is ​​a covariance matrix calculated based on the residuals of the operating conditions (usually multivariate residuals, i.e., residual vectors corresponding to multiple operating condition parameters). The overall fluctuation characteristic is a single value obtained by scalarizing the residual covariance matrix, used to comprehensively characterize the overall deviation or fluctuation intensity between the measured and estimated operating conditions within the current period. In this embodiment, the target residual threshold is an adaptive threshold limit calculated jointly based on the overall fluctuation characteristic and the risk critical coefficient. The risk critical coefficient is a pre-set proportional coefficient based on engineering experience, equipment importance, safety requirements, etc., used to quantify and calibrate the risk tolerance boundary.

[0082] In the specific implementation, firstly, a residual covariance matrix is ​​constructed based on the time-series residual sequence of operating conditions (i.e., residuals of multiple operating conditions at consecutive time points). Secondly, the sum of the main diagonal elements of the residual covariance matrix (the trace of the residual covariance matrix) is calculated, and after scalarization, the overall volatility characteristic is obtained. Finally, the overall volatility characteristic is jointly calculated with the risk threshold coefficient to obtain the target residual threshold, such as target residual threshold = overall volatility characteristic. 1 / 2 Risk threshold coefficient.

[0083] Optionally, to improve the statistical rigor and adaptability of the risk threshold coefficient setting, the risk threshold coefficient can be determined based on the distribution pattern of the overall fluctuation characteristics under historical normal operating conditions, using the chi-square distribution inverse cumulative distribution function, such as: Risk threshold coefficient = or risk threshold coefficient = ,in, The inverse cumulative distribution function of the chi-square distribution is... The pre-defined significance confidence parameter is S, which is the mean of the overall fluctuation characteristics under historical normal operating conditions.

[0084] In this embodiment, the above steps can be used to adapt the target residual threshold to different operating conditions, thereby improving the accuracy and fault tolerance of anomaly detection.

[0085] Furthermore, after step 215, the method further includes: if the absolute value of the residual of the operating condition is greater than the target residual threshold, then the number of anomalies is incremented by one, and it is determined whether the number of anomalies exceeds the preset number. If the number of anomalies does not exceed the preset number, then cross-modal feature fusion processing is triggered on the operating condition data sequence and equipment operation text data of the industrial equipment in the current cycle to obtain the fusion feature matrix of the industrial equipment.

[0086] Specifically, the number of anomalies is the statistical number of times the cumulative residual of the operating condition exceeds the target residual threshold. Its initial value is zero, and it is used to record the cumulative frequency of operating condition anomalies. The preset number is a threshold number pre-set based on the equipment's operational stability requirements and fault tolerance capabilities, such as 3 times or 5 times.

[0087] In practice, after determining whether the absolute value of the residual of the operating condition is greater than the target residual threshold, if it is, the number of anomalies is incremented by one, and it is determined whether the number of anomalies exceeds a preset number. If the number of anomalies does not exceed the preset number, it indicates that the current situation is only a momentary fluctuation or random disturbance, and has not yet formed a continuous operating condition anomaly. At this time, the step of cross-modal feature fusion processing of the operating condition data sequence and equipment operation text data of the industrial equipment in the current cycle can be triggered to obtain the fused feature matrix of the industrial equipment, that is, to carry out the control of the next cycle. If the number of anomalies exceeds the preset number, it indicates that the deviation of the equipment operating condition is a continuous, non-random disturbance, and it can be determined that the equipment has an operating abnormality or potential fault. At this time, the equipment abnormality alarm information can be generated and pushed to the terminal of the staff.

[0088] In this embodiment, the above steps can effectively avoid misjudgment problems caused by single random disturbances and instantaneous fluctuations.

[0089] Step 216: Determine that the operating status of the industrial equipment is abnormal, and send an alarm notification to the staff's terminal.

[0090] Specifically, alarm notifications are warning messages pushed to the terminals of maintenance personnel when the equipment is determined to be in an abnormal operating state. These messages are used to promptly inform the maintenance personnel of the abnormal equipment status, facilitating rapid manual verification and maintenance intervention.

[0091] In practice, once the absolute value of the residual of the operating condition is determined to be greater than the target residual threshold, the operating condition of the industrial equipment can be determined to be abnormal, and the industrial equipment can be judged to have been subjected to a replay attack. At this time, alarm prompts can be pushed to the terminals held by the staff (such as mobile phones, computers, etc.) so that the staff can promptly detect the safety anomaly and take appropriate intervention measures.

[0092] Step 217: Perform proxy optimization on the estimated operating condition data sequence to obtain the control strategy of the industrial equipment.

[0093] Specifically, surrogate optimization is a method of performing optimization calculations by using a surrogate model to replace a complex equipment mechanism model. A surrogate model is a low-computational-cost, relatively simple approximate mathematical model used to replace the high-fidelity but computationally complex and time-consuming original system model (such as an equipment mechanism model or a high-precision simulation model). Examples of surrogate models include polynomial response surfaces, radial basis function networks, Gaussian process regression, neural networks, support vector regression, and Kriging models.

[0094] In the specific implementation, the obtained estimated operating condition data sequence is input into a pre-built surrogate model. Then, a lightweight optimization algorithm (such as gradient descent, Bayesian optimization, differential evolution, etc.) is used to perform surrogate optimization on the estimated operating condition data sequence on the surrogate model to obtain the optimal control parameters of the industrial equipment. Specifically, candidate control variables are first initialized (the currently used control variables of the equipment can be selected as initial values); then, the surrogate model is called to predict the cost function value and constraint satisfaction degree corresponding to the current candidate control variables; then, the candidate control variables are iteratively updated according to the optimization algorithm; the model prediction and variable update steps are repeated until the objective function converges or the maximum number of iterations is reached to obtain the optimal control parameters of the industrial equipment. Finally, the corresponding equipment control strategy is matched according to the optimal control parameters of the industrial equipment. For example, if the optimal control parameter is a pulse width modulation duty cycle of 60% for the cooling fan, the corresponding equipment control strategy is to adjust the fan speed to that duty cycle.

[0095] For example, the construction process of the proxy model can be as follows: 1. Construct a working condition sample dataset: Collect working condition data sequences of the entire historical operation process of industrial equipment, and match them with corresponding control parameters and equipment operating performance index data. After data cleaning, outlier removal, and normalization preprocessing, a standard offline sample dataset for model training is formed. 2. Build and train a multi-base learner model: Select various machine learning models with nonlinear fitting and uncertainty assessment capabilities (such as deep neural networks, Gaussian process regression, support vector regression, etc.) as base learners. Using the preprocessed offline working condition sample dataset, train each base learner independently so that each base learner can individually fit the mapping relationship between equipment working condition data and operating performance; at the same time, obtain the prediction variance of each base learner under different working condition sample points through the model's own characteristics or statistical estimation methods to characterize the uncertainty of the single model prediction result. 3. Calculate the dynamic weights of the base learners: Based on the prediction variance of each base learner's output, and according to the preset weight mapping rules, quantify and calculate the dynamic weights corresponding to each base learner for each working condition sample point. Following the principle that the smaller the prediction variance, the higher the model's reliability, and the larger the assigned weights, adaptive allocation of the reliability of each base learner is achieved. 4. Weighted fusion to construct an integrated surrogate model: Combining the prediction outputs of each base learner with their corresponding dynamic weights, a weighted fusion calculation is performed. The prediction results of multiple base learners are collaboratively aggregated to form an integrated surrogate model that balances global fitting accuracy and working condition adaptability. The model structure, internal parameters, and weight calculation rules are then permanently saved, eliminating the need for real-time reconstruction and retraining. The aggregated prediction function formula can be expressed as: ; For the completed surrogate model, let represent the predictive control parameter result corresponding to the working condition input x; M is the total number of base learners; i is the index of the base learner; Let be the predicted output of the i-th base learner for the working condition input x; Let be the dynamic weight of the i-th base learner at the input x. The formula for calculating the dynamic weight can be expressed as: ;in, This is the weighting adjustment coefficient. Let be the prediction variance of the i-th base learner for the working condition input x; exp() is the exponential function. 5. Model Validation and Finalization Archiving: The accuracy, generalization ability, and robustness of the offline-built integrated surrogate model are verified using a test sample set. After passing the verification, the overall model parameters and weight algorithm logic are archived and solidified for direct use in the subsequent real-time online control stage of industrial equipment.

[0096] Optionally, during the offline construction phase of the surrogate model, the weight adjustment coefficients can be tuned offline. The specific steps are as follows: 1. Data partitioning: Divide the historical operation dataset of industrial equipment (such as the historical fusion feature matrix and its corresponding control parameters and operating performance indicators) into a training set (70%), a validation set (20%), and a test set (10%); 2. Initial training and evaluation: Set the initial value of the weight adjustment coefficients to 1, complete the training of each base learner on the training set, and calculate the prediction mean square error of the ensemble surrogate model based on the validation set; 3. Grid search search Advantages: 1. Employ a grid search method, iterating through the weight adjustment coefficients within a preset range [0.1, 100] with a step size of 0.1, and calculating the mean squared error of prediction on the validation set for each value; 2. Determine the optimal initial value: Select the weight adjustment coefficient corresponding to the minimum mean squared error of prediction on the validation set as the optimal initial value; 3. Verify generalization ability: Verify the model's generalization ability on the test set. If the mean squared error of prediction on the test set deviates from the mean squared error of prediction on the validation set by more than 15%, readjust the search range and execute steps 3-4 until the model's generalization performance meets the requirements.

[0097] In addition, an adaptive adjustment mechanism can be configured during the online operation of the surrogate model: if the prediction deviation of 100 consecutive sampling points exceeds the preset threshold, the online adaptive adjustment of the weight adjustment coefficient is triggered, with an adjustment step size of ±10% of the initial value, until the prediction deviation returns to the threshold range, so as to ensure the prediction stability of the surrogate model under different operating conditions.

[0098] In this embodiment, the above steps can effectively reduce computational complexity and quickly generate control strategies that adapt to the current operating conditions; while meeting the safety constraints of equipment operation, the operating efficiency can be optimized, adapting to the real-time control requirements of industrial equipment, and improving the optimization accuracy and operational robustness of the control strategy.

[0099] Step 218: Control industrial equipment based on control strategies.

[0100] The industrial equipment control method based on multi-source heterogeneous data provided in this invention first performs semantic quantization mapping on the equipment operation text data of the industrial equipment in the current cycle to obtain text semantic feature vectors. This transforms unstructured text data into structured feature representations, facilitating subsequent cross-modal fusion calculations with operating condition time-series data. Simultaneously, it overcomes the shortcomings of traditional solutions that rely solely on numerical operating condition data and ignore implicit semantic information in the text. This method can more comprehensively and completely depict the overall operating state of the industrial equipment, thereby improving the accuracy of subsequent equipment control decisions. Constructing an operating condition time-series matrix based on the operating condition data sequence of the industrial equipment in the current cycle can completely represent the temporal evolution of various operating parameters of the industrial equipment, improving the completeness and refinement of the equipment operating state representation, thus enhancing the accuracy of equipment control decisions. This paper utilizes a target attention mechanism to perform cross-modal attention fusion calculation on textual semantic feature vectors and operating condition time-series matrices, obtaining a fused feature matrix for industrial equipment. This breaks down the information barriers between textual semantic data and operating condition time-series data, integrating textual maintenance information and numerical operating parameters. It avoids the problem of one-sided representation by single-modal features, enabling the fused feature matrix to more comprehensively and accurately reproduce the true operating state of industrial equipment, thereby improving the accuracy and intelligence level of industrial equipment control decisions. Operating condition observation and calculation are performed on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment. This outputs accurate and continuous estimated operating condition data sequences, providing reliable time-series data support for subsequent industrial equipment control strategy formulation, thereby improving the accuracy and reliability of equipment control and ensuring the safe and stable operation of industrial equipment. Based on the operating condition data sequence and the estimated operating condition data sequence, the operating condition residual is calculated, which can quantitatively represent the deviation between the state estimation result and the actual operating state of the equipment. This provides accuracy feedback for subsequent closed-loop control and reliable quantitative judgment indicators for operating condition anomaly identification and fault early warning. Finally, it determines whether the absolute value of the operating condition residual is greater than the target residual threshold. If the value is greater than the specified value, the industrial equipment's operating status is determined to be abnormal, and an alarm message is sent to the operator's terminal so that the operator can promptly detect the safety anomaly and take appropriate intervention measures. If the value is not greater than the specified value, the estimated operating condition data sequence is processed by proxy optimization to obtain the control strategy for the industrial equipment. This effectively reduces computational complexity and quickly generates a control strategy adapted to the current operating conditions. While meeting the equipment's operational safety constraints, it optimizes operational efficiency, adapts to the real-time control needs of industrial equipment, and improves the optimization accuracy and operational robustness of the control strategy. Controlling industrial equipment based on the control strategy improves the timeliness and intelligence of equipment control, ensures the stable and efficient operation of industrial equipment, and enhances the accuracy and practicality of overall control. Therefore, the technical solution of this invention overcomes the shortcomings of existing technologies that rely solely on measured operating condition data, making it difficult to comprehensively assess the overall operating status of the equipment, which can easily lead to one-sided control decisions and insufficient accuracy, thus failing to meet the needs of refined and intelligent control of industrial equipment.

[0101] Figure 3 This is a schematic diagram of an industrial equipment control device based on multi-source heterogeneous data provided in an embodiment of the present invention. This device and the industrial equipment control method based on multi-source heterogeneous data in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the industrial equipment control device based on multi-source heterogeneous data, please refer to the embodiments of the industrial equipment control method based on multi-source heterogeneous data described above.

[0102] like Figure 3 As shown, the device includes: The fusion module 310 is used to perform cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment. The calculation module 320 is used to perform operating condition observation calculation on the fused feature matrix to obtain the estimated operating condition data sequence of the industrial equipment; The control module 330 is used to determine the control strategy of the industrial equipment based on the operating condition data sequence and the estimated operating condition data sequence, and to control the industrial equipment based on the control strategy.

[0103] Based on the above embodiments, the fusion module 310 performs cross-modal feature fusion processing on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment, including: Semantic quantization mapping is performed on the equipment operation text data of industrial equipment in the current period to obtain text semantic feature vectors; a working condition time series matrix is ​​constructed based on the working condition data sequence of industrial equipment in the current period; cross-modal attention fusion calculation is performed on the text semantic feature vectors and the working condition time series matrix using a target attention mechanism to obtain the fusion feature matrix of the industrial equipment.

[0104] Based on the above embodiments, the query matrix of the target attention mechanism is the working condition time series matrix, the key matrix of the target attention mechanism is obtained by extending the text semantic feature vector, and the value matrix of the target attention mechanism is obtained by extending the text semantic feature vector.

[0105] Based on the above embodiments, the fusion module 310 performs semantic quantization mapping processing on the equipment operation text data of industrial equipment in the current cycle to obtain a text semantic feature vector, including: The semantic items in the device's running text data are determined; the fuzzy semantic quantization parameters corresponding to the semantic items are determined based on preset fuzzy semantic mapping rules; a preset number of random simulated samples are generated based on the fuzzy semantic quantization parameters to obtain a fuzzy feature simulation dataset of the semantic items; and feature extraction is performed on the fuzzy feature simulation dataset to obtain the text semantic feature vector.

[0106] Based on the above embodiments, the calculation module 320 is specifically used for: The current operating condition of the industrial equipment is determined; based on the current operating condition, a matching is performed in a preset operating condition observation parameter library to obtain a set of state estimation parameters for the current operating condition; the fused feature matrix and the set of state estimation parameters are substituted into a preset estimation calculation model to obtain an estimated operating condition data sequence for the industrial equipment.

[0107] Based on the above embodiments, the control module 330 determines the control strategy of the industrial equipment according to the operating condition data sequence and the estimated operating condition data sequence, including: Based on the operating condition data sequence and the estimated operating condition data sequence, the operating condition state residual is calculated; it is determined whether the absolute value of the operating condition state residual is greater than the target residual threshold; if the absolute value of the operating condition state residual is not greater than the target residual threshold, then the estimated operating condition data sequence is subjected to surrogate optimization solution processing to obtain the control strategy of the industrial equipment.

[0108] Based on the above embodiments, the device further includes: The threshold determination module is used to construct a residual covariance matrix based on the operating condition residual before determining whether the absolute value of the residual of the operating condition is greater than the target residual threshold; calculate the trace of the residual covariance matrix to obtain the overall volatility characteristic; and calculate the target residual threshold based on the overall volatility characteristic and the risk critical coefficient.

[0109] Based on the above embodiments, the device further includes: The early warning module is used to determine whether the absolute value of the residual of the operating condition is greater than the target residual threshold. If the absolute value of the residual of the operating condition is greater than the target residual threshold, the module determines that the operating condition of the industrial equipment is abnormal and sends an alarm message to the operator's terminal.

[0110] Based on the above embodiments, the device further includes: The judgment module is used to determine whether the absolute value of the residual of the operating condition is greater than the target residual threshold. If the absolute value of the residual of the operating condition is greater than the target residual threshold, the abnormality count is incremented by one, and it is determined whether the abnormality count exceeds a preset number. If the abnormality count does not exceed the preset number, cross-modal feature fusion processing is triggered on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment.

[0111] The industrial equipment control device based on multi-source heterogeneous data provided in the embodiments of the present invention can execute the industrial equipment control method based on multi-source heterogeneous data provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0112] It is worth noting that in the above embodiments of the industrial equipment control device based on multi-source heterogeneous data, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0113] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 4 A block diagram of an exemplary electronic device 4 suitable for implementing embodiments of the present invention is shown. Figure 4 The electronic device 4 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0114] like Figure 4 As shown, electronic device 4 is represented in the form of a general-purpose computing electronic device. The components of electronic device 4 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0115] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0116] Electronic device 4 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 4, including volatile and non-volatile media, removable and non-removable media.

[0117] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Electronic device 4 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (… Figure 4 Not shown; usually referred to as a "hard drive"). Although Figure 4 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0118] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0119] Electronic device 4 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with electronic device 4, and / or with any device that enables electronic device 4 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, electronic device 4 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. Figure 4 As shown, network adapter 20 communicates with other modules of electronic device 4 via bus 18. It should be understood that, although... Figure 4 Not shown, it can be combined with electronic device 4 to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0120] The processing unit 16 executes various functional applications and page displays by running programs stored in the system memory 28, such as implementing the industrial equipment control method based on multi-source heterogeneous data provided in the embodiments of the present invention.

[0121] Of course, those skilled in the art will understand that the processor can also implement the technical solution of the industrial equipment control method based on multi-source heterogeneous data provided in any embodiment of the present invention.

[0122] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements, for example, the industrial equipment control method based on multi-source heterogeneous data provided in this invention.

[0123] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0125] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0126] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0127] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0128] Furthermore, the acquisition, storage, use, and processing of data in the technical solution of this invention all comply with relevant laws and regulations.

[0129] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. An industrial equipment control method based on multi-source heterogeneous data, characterized in that, The method includes: Cross-modal feature fusion processing is performed on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment. The fused feature matrix is ​​subjected to operating condition observation and calculation to obtain the estimated operating condition data sequence of the industrial equipment; The control strategy for the industrial equipment is determined based on the operating condition data sequence and the estimated operating condition data sequence, and the industrial equipment is controlled based on the control strategy.

2. The method according to claim 1, characterized in that, Cross-modal feature fusion processing is performed on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fused feature matrix of the industrial equipment, including: Semantic quantization mapping is performed on the equipment operation text data of industrial equipment in the current cycle to obtain text semantic feature vectors; Construct a time series matrix of operating conditions based on the operating condition data sequence of industrial equipment within the current period; The text semantic feature vector and the working condition time series matrix are subjected to cross-modal attention fusion calculation using a target attention mechanism to obtain the fusion feature matrix of the industrial equipment.

3. The method according to claim 2, characterized in that, The query matrix of the target attention mechanism is the working condition time series matrix, the key matrix of the target attention mechanism is obtained by extending the text semantic feature vector, and the value matrix of the target attention mechanism is obtained by extending the text semantic feature vector.

4. The method according to claim 2, characterized in that, Semantic quantization mapping is performed on the text data of industrial equipment operation within the current period to obtain text semantic feature vectors, including: Determine the semantic items in the device's running text data; The fuzzy semantic quantization parameters corresponding to the semantic item are determined based on a preset fuzzy semantic mapping rule; Based on the fuzzy semantic quantization parameters, a preset number of random simulated samples are generated to obtain the fuzzy feature simulation dataset of the semantic item; Feature extraction is performed on the fuzzy feature simulation dataset to obtain the text semantic feature vector.

5. The method according to claim 1, characterized in that, The fused feature matrix is ​​subjected to operating condition observation and solution to obtain the estimated operating condition data sequence of the industrial equipment, including: Determine the current operating condition of the industrial equipment; Based on the current operating conditions, a set of state estimation parameters for the current operating conditions is obtained by matching them in a preset operating condition observation parameter library. Substituting the fused feature matrix and the state estimation parameter set into a preset estimation calculation model, the estimated operating condition data sequence of the industrial equipment is obtained.

6. The method according to claim 1, characterized in that, Determining the control strategy for the industrial equipment based on the operating condition data sequence and the estimated operating condition data sequence includes: Based on the operating condition data sequence and the estimated operating condition data sequence, calculate the operating condition residual; Determine whether the absolute value of the residual under the operating condition is greater than the target residual threshold; If the absolute value of the residual of the operating condition is not greater than the target residual threshold, then the estimated operating condition data sequence is subjected to surrogate optimization solution processing to obtain the control strategy of the industrial equipment.

7. The method according to claim 6, characterized in that, Before determining whether the absolute value of the residual under the operating condition is greater than the target residual threshold, the method further includes: Construct a residual covariance matrix based on the residuals of the aforementioned operating conditions; The trace of the residual covariance matrix is ​​calculated to obtain the overall fluctuation characteristic. The target residual threshold is calculated based on the overall volatility characteristic and the risk critical coefficient.

8. The method according to claim 6, characterized in that, After determining whether the absolute value of the residual under the operating condition is greater than the target residual threshold, the method further includes: If the absolute value of the residual of the operating condition is greater than the target residual threshold, the operating condition of the industrial equipment is determined to be abnormal, and an alarm prompt message is sent to the operator's terminal.

9. The method according to claim 6, characterized in that, After determining whether the absolute value of the residual under the operating condition is greater than the target residual threshold, the method further includes: If the absolute value of the residual of the operating condition is greater than the target residual threshold, the number of anomalies is incremented by one, and it is determined whether the number of anomalies exceeds the preset number. If the number of anomalies does not exceed the preset number, cross-modal feature fusion processing is triggered on the operating condition data sequence and equipment operation text data of the industrial equipment in the current period to obtain the fusion feature matrix of the industrial equipment.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the industrial equipment control method based on multi-source heterogeneous data as described in any one of claims 1-9.

Images