Big model-based data processing method and device and computer equipment

Abstract

The application relates to a large model-based data processing method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring device data collected by a production device and a prompt word template matched with the production device; the prompt word template comprises a first variable placeholder and a second variable placeholder; according to the data type of the device data, first sub-information matched with the device data is generated, and the first sub-information is filled into the first variable placeholder; a processing intention for the device data is determined, second sub-information matched with the processing intention is generated, and the second sub-information is filled into the second variable placeholder; the prompt word template after the filling information is used as prompt information, and a large language model is guided to process the device data through the prompt information, so that a data processing result matched with the processing intention is obtained. The method can improve the efficiency and accuracy of data processing.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on a large model. Background Technology

[0002] With the deepening development of the Industrial Internet and intelligent manufacturing, large-scale sensor networks can be deployed to collect different types of data from production equipment. This data is a key foundation for realizing equipment status monitoring, fault diagnosis, predictive maintenance, and full lifecycle management.

[0003] However, for different types of data collected from production equipment, the data is usually verified manually, which makes the data processing efficiency and accuracy low. Summary of the Invention

[0004] Therefore, it is necessary to provide a data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on a large model that can improve data processing efficiency and accuracy in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a data processing method based on a large model, including:

[0006] Acquire equipment data collected for the production equipment, and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder;

[0007] Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder;

[0008] Determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder;

[0009] The prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain a data processing result that matches the processing intention.

[0010] In one embodiment, the processing intent includes correcting abnormal data in the device data; the step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes:

[0011] Obtain the preset data range corresponding to the data type of the production equipment;

[0012] The prompt information and the device data are input into the large language model to obtain an initial result that matches the processing intent; the initial result includes sensor data corresponding to multiple time points.

[0013] If there is abnormal data in the sensor data that does not match the preset data range, the abnormal data is corrected based on the preset data range to obtain updated data;

[0014] Based on the updated data, a data processing result matching the processing intent is obtained.

[0015] In one embodiment, obtaining a data processing result matching the processing intent based on each of the updated data includes:

[0016] Combined data is obtained by combining the updated data and the normal data from each of the sensor data that matches the preset data range;

[0017] Obtain feedback information regarding the combined data;

[0018] If the feedback information verification is successful, the combined data will be determined as the data processing result that matches the processing intention.

[0019] In one embodiment, the processing intent includes correcting abnormal data in the device data; the step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes:

[0020] Obtain historical normal data and historical abnormal data of the production equipment under the data type described above;

[0021] The prompts guide the large language model to identify and correct abnormal data in the device data by referring to the differences between the historical normal data and the historical abnormal data, thereby obtaining a data processing result that matches the processing intent.

[0022] In one embodiment, the device data includes multiple sub-data corresponding to different times, and the processing intent includes determining the data location of abnormal data in the device data based on the data change amplitude between adjacent time sub-data in the device data;

[0023] The step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes:

[0024] The prompt information guides the large language model to output the location of abnormal data in the device data based on the magnitude of data change between adjacent time-series sub-data in the device data.

[0025] Correct the sub-data corresponding to the abnormal data location to obtain a data processing result that matches the processing intention.

[0026] In one embodiment, correcting the sub-data corresponding to the abnormal data location to obtain a data processing result matching the processing intent includes:

[0027] Determine the data change trend of the device data, generate third sub-information that matches the data change trend, and fill the third sub-information into the prompt information;

[0028] The prompt information after filling in the information guides the large language model to correct the sub-data corresponding to the abnormal data location based on the data range represented by the common trend in the data change trend, so as to obtain a data processing result that matches the processing intention.

[0029] Secondly, this application also provides a data processing apparatus based on a large model, comprising:

[0030] The acquisition module is used to acquire equipment data collected from the production equipment and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder;

[0031] The first processing module is configured to generate first sub-information matching the device data according to the data type of the device data, and fill the first sub-information into the first variable placeholder;

[0032] The second processing module is used to determine the processing intention for the device data, generate second sub-information that matches the processing intention, and fill the second sub-information into the second variable placeholder.

[0033] The third processing module is used to take the prompt word template after filling in the information as prompt information, and guide the large language model to process the device data through the prompt information to obtain a data processing result that matches the processing intention.

[0034] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0035] Acquire equipment data collected for the production equipment, and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder;

[0036] Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder;

[0037] Determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder;

[0038] The prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain a data processing result that matches the processing intention.

[0039] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0040] Acquire equipment data collected for the production equipment, and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder;

[0041] Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder;

[0042] Determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder;

[0043] The prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain a data processing result that matches the processing intention.

[0044] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0045] Acquire equipment data collected for the production equipment, and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder;

[0046] Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder;

[0047] Determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder;

[0048] The prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain a data processing result that matches the processing intention.

[0049] The aforementioned data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on a large model acquire equipment data collected from production equipment and a prompt word template matching the production equipment. The prompt word template includes a first variable placeholder and a second variable placeholder. This step, by pre-constructing a prompt word template matching the production equipment, replaces the tedious manual process of redesigning the verification process each time, significantly shortening the startup time from data acquisition to processing and improving data processing efficiency. By generating first sub-information matching the equipment data based on its data type and filling it into the first variable placeholder, and by determining the processing intent for the equipment data and generating second sub-information matching the processing intent and filling it into the second variable placeholder, the raw, heterogeneous equipment data and abstract processing targets are transformed into standardized semantic instructions that the large language model can accurately understand. This process replaces the two cognitive tasks of simultaneously "identifying the data type" and "determining the processing method" required in traditional manual verification, avoiding classification errors and intent biases caused by human fatigue or lack of experience, and thus improving the accuracy of data processing. By using the prompt word template after filling in the information as prompt information, and guiding the large language model to process the device data through the prompt information, the data processing results that match the processing intent are obtained. Thus, the large language model can automatically execute data analysis tasks that meet the expectations, significantly improving the automation, accuracy and overall efficiency of data processing. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating a data processing method based on a large model in one embodiment;

[0052] Figure 2This is a flowchart illustrating how a large language model is guided by prompts to process device data in one embodiment, resulting in a data processing outcome that matches the processing intent.

[0053] Figure 3 This is a flowchart illustrating how a large language model is guided to process device data through prompts in another embodiment, resulting in a data processing outcome that matches the processing intent.

[0054] Figure 4 This is a flowchart illustrating a data processing method based on a large model in another embodiment;

[0055] Figure 5 This is a structural block diagram of a data processing device based on a large model in one embodiment;

[0056] Figure 6 This is an internal structural diagram of a computer device in one embodiment;

[0057] Figure 7 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various variable placeholders, but these variable placeholders are not limited by these terms. These terms are only used to distinguish between the first variable placeholder and the second variable placeholder. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the schemes, or any combination of multiple schemes.

[0060] In real-world scenarios, sensor networks collect equipment data (such as massive amounts of time-series data like temperature, pressure, vibration frequency, current, and voltage) from production equipment. This data forms the core foundation for building equipment early warning models and digital twin models; therefore, data quality directly determines the model's analytical accuracy and early warning effectiveness. However, in actual data collection, equipment data often faces two key challenges:

[0061] The problem of missing values ​​refers to the large number of continuous or discrete missing points in time-series data due to sensor failures, signal transmission interruptions, data acquisition system lags, etc. However, directly deleting missing data will cause the time series to break and fail to reflect the complete operating status of the equipment; simple mean imputation, linear interpolation and other methods ignore the long-term trends, seasonal fluctuations and periodic patterns of industrial data, and the imputation results lack physical meaning and are prone to introducing systematic errors.

[0062] Interference noise is a problem caused by factors such as electromagnetic interference in the field and sudden abnormal start-up and shutdown of equipment, which can lead to abnormal spikes and drifts in equipment data. If this noise is not effectively eliminated, it can cause false alarms or missed alarms in the model's learning of normal equipment operating conditions based on the severe interference early warning model, thus reducing the simulation reliability of the digital twin model.

[0063] Traditional technologies can preprocess missing or outlier data manually, but this results in low data processing efficiency. Alternatively, traditional statistical methods (such as moving averages and Kalman filtering) can be used for data preprocessing, but these methods are poorly adapted to complex, nonlinear, and non-stationary industrial data, and struggle to accurately capture the deep, implicit features of the data. Machine learning algorithms can be used for missing value imputation, but this still relies on manual feature engineering and is inefficient when processing large-scale, multi-dimensional equipment data, failing to meet the real-time and intelligent requirements of modern industry. In other words, traditional data preprocessing methods (such as mean imputation and linear interpolation) cannot effectively capture the long-term trends, seasonality, and periodicity of industrial equipment data, leading to unreasonable missing value imputation results and compromising data consistency and physical authenticity. Furthermore, they struggle to effectively identify and clean complex noise and outliers in equipment data; residual noise severely reduces the training accuracy and predictive reliability of subsequent early warning models and digital twin models. Moreover, the lack of automated feature extraction and dynamic update mechanisms for time-series data makes it difficult for early warning models and digital twin models to adapt to the dynamic changes in equipment operating data over time, resulting in performance degradation over time. Therefore, there is an urgent need for an intelligent data preprocessing solution that can automatically extract historical data features, accurately fill in missing values, and effectively clean up interference noise, in order to meet the application requirements of high-precision early warning models and digital twin models.

[0064] In view of this, this application provides a data processing method based on a large model to achieve high-quality missing value imputation and noise removal through precise data analysis, outputting highly complete and accurate time-series data to support the efficient training and stable operation of subsequent models. Specifically, the powerful semantic understanding and sequence generation capabilities of large language models can be utilized to perform in-depth analysis, feature mining, missing value imputation, and noise removal on the collected device data, and to achieve dynamic iterative optimization of the model.

[0065] The large-model-based data processing method provided in this application can be applied to computing devices, which can be terminals or servers. Terminals can be, but are not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart vehicle devices, projection devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted displays, etc. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Servers can be independent physical servers, server clusters or distributed systems composed of multiple physical servers, or cloud servers providing cloud computing services.

[0066] In one exemplary embodiment, such as Figure 1 As shown, a data processing method based on a large model is provided. Taking the application of this method to a computer device as an example, the method includes the following steps:

[0067] S102, acquire equipment data collected for the production equipment and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder.

[0068] Production equipment refers to physical entities that participate in product manufacturing, energy conversion, or material handling processes, capable of independently or collaboratively completing specific production operations. This includes, but is not limited to, power generating units, chemical plants, and mechanical production lines. Equipment data refers to data collected through a sensor network deployed on the production equipment. The sensor network integrates various types of sensors to comprehensively collect different operating parameters of the production equipment according to a data sampling frequency; the data sampling frequency can be 1 second / time or other values. For example, equipment data includes multiple sub-data points corresponding to different moments, including but not limited to operating parameters such as main steam temperature, furnace pressure, turbine vibration, or generator current.

[0069] A cue word template is a pre-designed, structured text containing variable placeholders, used to guide a large language model in generating answers that meet specific needs. For example, a cue word template includes a first variable placeholder and a second variable placeholder. The first variable placeholder is a pre-defined marker in the cue word template that indicates the data type of the data to be entered for analysis. The second variable placeholder is a pre-defined marker in the cue word template that indicates the processing intent of the data to be entered for analysis.

[0070] For example, the content of the prompt word template can be described as "analyzing the long-term trend, seasonal variation pattern, and periodic pattern of {first variable placeholder}, and making a comprehensive judgment based on the normal operating conditions of the equipment and historical data characteristics of similar time periods to realize {second variable placeholder}". Thus, guided by the prompt word template, the large language model can output data processing results that conform to the processing intent by using the identified historical features and contextual information.

[0071] In an optional embodiment, obtaining a prompt word template matching the production equipment includes: matching the equipment identifier of the production equipment with the preset mapping relationship between equipment identifiers and prompt word templates to determine the prompt word template matching the production equipment. Different equipment identifiers can correspond to the same prompt word template. Thus, automatically matching preset prompt word templates with equipment identifiers replaces the tedious process of manually writing verification rules or prompt words for each device, avoiding manual configuration errors and inefficiency caused by the diversity and large number of equipment types, and helping to improve data processing efficiency.

[0072] S104, Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder.

[0073] Here, data type refers to the category of physical quantity described by the equipment data, including but not limited to: temperature, pressure, vibration, voltage, and current. The first sub-information refers to the specific content generated based on the data type of the equipment data, used to fill the first variable placeholder in the prompt word template. For example, taking temperature as the data type, the content of the first sub-information can be described as "temperature-type time-series data of the thermodynamic state of the production equipment"; taking pressure as the data type, the content of the first sub-information can be described as "pressure-type time-series data of the mechanical state of the production equipment".

[0074] S106, determine the processing intention for the device data, generate second sub-information that matches the processing intention, and fill the second sub-information into the second variable placeholder.

[0075] The processing intent refers to the analytical operation objectives that are expected to be performed on the device data, including but not limited to: missing data location detection, missing data correction, abnormal data correction, data trend prediction, and data status assessment.

[0076] In an optional embodiment, determining the processing intent for device data includes: responding to an input operation in response to a graphical user interface; and determining the input content corresponding to the input operation as the processing intent for device data. Thus, by allowing users to flexibly specify the analysis targets for device data through simple and intuitive interface operations, it ensures that the output results of subsequent data processing are more closely aligned with the user's actual needs, thereby helping to improve the accuracy of data processing and user satisfaction.

[0077] The second sub-information refers to the specific content generated based on the processing intent of the device data and used to fill the second variable placeholder in the prompt word template. It is a clear statement of the analysis operation objective. For example, if the processing intent is to correct abnormal data, the content of the second sub-information can be described as "identify and correct outliers in the data"; if the processing intent is to detect missing data locations, the content of the second sub-information can be described as "identify missing locations in the data".

[0078] S108, the prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain the data processing result that matches the processing intention.

[0079] Using the method described in the above embodiments, equipment data collected from production equipment and a prompt word template matching the production equipment are obtained. The prompt word template includes a first variable placeholder and a second variable placeholder. This step, by pre-constructing a prompt word template matching the production equipment, replaces the tedious manual process of redesigning the verification process each time, significantly shortening the startup time from data collection to processing and improving data processing efficiency. By generating first sub-information matching the equipment data based on its data type and filling it into the first variable placeholder, and by determining the processing intent for the equipment data, generating second sub-information matching the processing intent and filling it into the second variable placeholder, the raw, heterogeneous equipment data and abstract processing targets are transformed into standardized semantic instructions that a large language model can accurately understand. This process replaces the traditional manual verification process, which requires simultaneously completing the cognitive tasks of "identifying the data type" and "determining the processing method," avoiding classification errors and intent biases caused by human fatigue or lack of experience, and thus improving the accuracy of data processing. By using the prompt word template after filling in the information as prompt information, and guiding the large language model to process the device data through the prompt information, the data processing results that match the processing intent are obtained. Thus, the large language model can automatically execute data analysis tasks that meet the expectations, significantly improving the automation, accuracy and overall efficiency of data processing.

[0080] In one embodiment, the method further includes: combining multiple sets of data processing results to obtain a time-series dataset; and updating the early warning model or digital twin model corresponding to the production equipment based on the time-series dataset. Thus, by continuously updating the parameters and feature weights of the model using the time-series dataset as training samples, the early warning model and digital twin model can dynamically adapt to changes in equipment operating data over time, maintaining good analytical and predictive performance over the long term.

[0081] In one embodiment, the processing intent includes correcting anomalous data in the device data. Anomalous data may include missing data and outlier data, where outlier data refers to data that deviates from the normal fluctuation range. For example... Figure 2 As shown, the large language model is guided by prompts to process device data and obtain data processing results that match the processing intent. This includes the following steps:

[0082] S1081, Obtain the preset data range corresponding to the data type of the production equipment.

[0083] Among them, the preset data range refers to the normal value range of a physical quantity when the equipment is operating normally, which is pre-set for different data types of production equipment (such as temperature, pressure, vibration, etc.).

[0084] In an optional embodiment, obtaining the preset data range corresponding to the production equipment under a data type includes: matching the data type with the preset data type and normal data range under the production equipment to determine the preset data range corresponding to the production equipment under the data type. Thus, by pre-setting the mapping relationship, rapid matching of the data range with the production equipment and data type can be achieved, which helps improve the efficiency and accuracy of anomaly detection.

[0085] S1083, input the prompt information and device data into the large language model to obtain an initial result that matches the processing intent; the initial processing result includes sensor data corresponding to multiple time points.

[0086] S1085, if there is abnormal data in each sensor data that does not match the preset data range, correct the abnormal data based on the preset data range to obtain updated data.

[0087] In an optional embodiment, if the sensor data is outside a preset data range, the sensor data is identified as abnormal data.

[0088] In an optional embodiment, correcting abnormal data based on a preset data range to obtain updated data includes: modifying the abnormal data to any value within the preset data range to obtain updated data.

[0089] S1087, Based on each updated data, obtain the data processing result that matches the processing intention.

[0090] In an optional embodiment, based on each updated data, a data processing result matching the processing intention is obtained, including: determining each updated data and normal data in each sensor data that matches a preset data range as the data processing result matching the processing intention.

[0091] In an optional embodiment, based on each updated data, a data processing result matching the processing intention is obtained, including: obtaining combined data by combining each updated data and normal data from each sensor data that matches a preset data range; obtaining feedback information for the combined data; and determining the combined data as the data processing result matching the processing intention if the feedback information characterization verification passes.

[0092] Feedback information refers to the judgment information used to characterize whether the combined data has passed the quality verification. For example, feedback information includes a clear status indicator of "verification passed" or "verification failed".

[0093] In the above embodiments, by combining updated data with normal data and introducing feedback information for verification, a closed-loop mechanism combining automated correction and manual confirmation is formed. This ensures, on the one hand, the integrity and temporal continuity of the output data, avoiding the potential unreliability risks of directly generating results using large language models; on the other hand, by only determining the combined data as the final processing result after verification, it balances the efficient processing capabilities of large language models with the stringent accuracy requirements of industrial scenarios. Ultimately, this significantly improves the credibility, security, and practical usability of the abnormal data correction task, providing a high-quality data foundation for subsequent status monitoring and fault diagnosis.

[0094] In one embodiment, the processing intent includes correcting anomalous data in the equipment data. A large language model is guided by prompts to process the equipment data and obtain a data processing result matching the processing intent. This includes: obtaining historical normal data and historical anomalous data of the production equipment under different data types; and using the differences between the historical normal data and historical anomalous data as a reference, the large language model is guided by prompts to identify and correct anomalous data in the equipment data, thereby obtaining a data processing result matching the processing intent.

[0095] Historical normal data refers to records of normal operating parameters collected by sensor networks and verified within preset data ranges during historical periods when the production equipment was in a healthy operating state, without significant faults or abnormal operating conditions. Historical abnormal data refers to records of operating parameters that deviate from the normal physical operating patterns of the equipment, exceed preset data ranges, or exhibit typical distortion patterns such as spikes, drifts, jams, or missing data, caused by sensor failures, communication interference, sudden changes in operating conditions, or equipment deterioration during the historical operation of the production equipment.

[0096] Understandably, large language models can extract the feature differences between normal fluctuations and abnormal interference (such as spikes and drifts) in device data by comparing and learning from historical normal data and historical abnormal data. Based on these feature differences, abnormal data can be identified and corrected based on historical normal data.

[0097] In the above embodiments, historical normal data provides an operating benchmark for the health status of the equipment, while historical abnormal data presents typical distortion patterns (such as spikes, drifts, and lag). The comparison between the two enables the model to accurately capture the essential differences between normal fluctuations and abnormal interference. Thus, by guiding the large language model to learn the feature differences between historical normal data and historical abnormal data, it helps to improve the accuracy of abnormal data identification and correction.

[0098] In one embodiment, device data includes multiple sub-data corresponding to different times, and the processing intent includes determining the location of abnormal data in the device data based on the magnitude of data change between adjacent sub-data at different times. For example... Figure 3 As shown, the large language model is guided by prompts to process device data and obtain data processing results that match the processing intent. This includes the following steps:

[0099] S1082 guides the large language model with prompts and outputs the location of abnormal data in the device data based on the magnitude of data change between adjacent time sub-data in the device data.

[0100] The location of abnormal data refers to the position of abnormal data in the device data, which can be represented by a time index.

[0101] In an optional embodiment, the large language model is guided by prompts to output the location of abnormal data in the device data based on the magnitude of data change between adjacent time sub-data in the device data. This includes: guiding the large language model with prompts to identify all adjacent time pairs in the device data whose magnitude of data change between adjacent time sub-data exceeds a magnitude threshold; and using the large language model to output the common time index in the adjacent time pairs as the location of abnormal data.

[0102] S1084, correct the sub-data corresponding to the abnormal data location to obtain the data processing result that matches the processing intention.

[0103] In an optional embodiment, the sub-data corresponding to the abnormal data location is corrected to obtain a data processing result matching the processing intent. This includes: based on a preset data range corresponding to the data type of the production equipment, the sub-data corresponding to the abnormal data location is corrected to obtain updated data; and each updated data, as well as the sub-data corresponding to the data locations other than the abnormal data location in the equipment data, are determined as the data processing result matching the processing intent. Thus, by using a preset data range to reasonably replace abnormal data, the efficiency of abnormal data processing is significantly improved while ensuring the quality and reliability of the output data, providing a high-quality data foundation for subsequent equipment status monitoring and fault analysis.

[0104] In an optional embodiment, correcting the sub-data corresponding to the abnormal data location to obtain a data processing result matching the processing intent includes: determining the data change trend of the device data, generating third sub-information matching the data change trend, and filling the third sub-information into the prompt information; guiding the large language model to correct the sub-data corresponding to the abnormal data location using the reference data range represented by the common trend in the data change trend through the prompt information after filling the information, so as to obtain a data processing result matching the processing intent.

[0105] The reference data range refers to the range of values ​​extracted from the time series of equipment data through statistical analysis of multiple normal variation segments or periodic patterns, which can represent the normal fluctuation boundary of the equipment's data under healthy operating conditions.

[0106] In an optional embodiment, the prompt information after filling in the information guides the large language model to correct the sub-data corresponding to the abnormal data position based on the data range represented by the common trend in the data change trend, so as to obtain a data processing result that matches the processing intention. This includes: guiding the large language model to correct the sub-data corresponding to the abnormal data position based on the data range represented by the common trend in the data change trend based on the prompt information after filling in the information, so as to obtain updated data; and guiding the large language model to combine each updated data and the sub-data corresponding to the data positions other than the abnormal data position in the device data based on the prompt information after filling in the information, so as to obtain a data processing result that matches the processing intention.

[0107] In the above embodiments, by dynamically injecting the data change trend as a third sub-information into the prompt information, the large language model is guided to use the reference data range represented by the common trend as a correction benchmark. This allows the model to extract dynamic boundaries from the data's own evolutionary patterns, helping to adapt to normal fluctuations under different operating conditions. Since the common trend range reflects the true change patterns in the equipment's health state, correcting outliers in this way ensures that the correction results conform to the inherent dynamic characteristics of the equipment, avoiding over-correction or under-correction. Ultimately, this achieves a technological leap from fixed rules to trend-adaptive anomaly data correction, significantly improving the rationality of the correction results and its adaptability to industrial scenarios.

[0108] In summary, such as Figure 4 As shown, a data processing method based on a large model is provided. Taking the application of this method to a computer device as an example, the method includes the following steps:

[0109] S401, acquire equipment data collected for the production equipment and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder.

[0110] S402, based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder.

[0111] S403, determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder; the processing intent includes correcting abnormal data in the device data.

[0112] S404, Determine the preset data range corresponding to the data type of the production equipment.

[0113] S405, input the prompt information and device data into the large language model to obtain the initial processing result that matches the processing intent; the initial processing result includes sensor data corresponding to multiple time points.

[0114] S406, If there is abnormal data in each sensor data that does not match the preset data range, correct the abnormal data based on the preset data range to obtain updated data.

[0115] S407, combined data is obtained by combining each updated data and the normal data in each sensor data that matches the preset data range.

[0116] S408, obtain feedback information for the combined data.

[0117] S409: If the feedback information characterization verification passes, the combined data will be determined as the data processing result that matches the processing intention.

[0118] In an optional embodiment, the processing intent includes correcting abnormal data in the equipment data; guiding the large language model to process the equipment data through prompts to obtain data processing results that match the processing intent, including: obtaining historical normal data and historical abnormal data of the production equipment under the data type; guiding the large language model to identify and correct abnormal data in the equipment data with reference to the differences between historical normal data and historical abnormal data through prompts to obtain data processing results that match the processing intent.

[0119] In an optional embodiment, the device data includes multiple sub-data corresponding to different times. The processing intent includes determining the location of abnormal data in the device data based on the data change amplitude between adjacent sub-data at different times; guiding a large language model to process the device data through prompts to obtain a data processing result matching the processing intent, including: guiding the large language model through prompts to output the location of abnormal data in the device data based on the data change amplitude between adjacent sub-data at different times; determining the data change trend of the device data, generating third sub-information matching the data change trend, and filling the prompts with the third sub-information; and guiding the large language model through the prompts after filling the information to correct the sub-data corresponding to the location of abnormal data based on the data range represented by the common trend in the data change trend, thereby obtaining a data processing result matching the processing intent.

[0120] The specific content of S401-S409 can be found in the aforementioned description and will not be repeated here.

[0121] Based on the above, taking a thermal power generating unit as an example, this unit is equipped with multiple sensors that collect key parameters in real time, such as main steam temperature, furnace pressure, turbine vibration, and generator current, with a data sampling frequency of 1 second per sampling. During long-term operation, some sensors experience signal interruptions due to high temperatures and vibrations, resulting in data gaps. Simultaneously, electromagnetic interference in the field often causes data spikes and noise, severely affecting model accuracy. Therefore, the data processing method provided in this application can be used for the data collected from the thermal power generating unit. Specifically:

[0122] Obtain the main steam temperature time-series data sequence (i.e., equipment data) for the unit over the past 7 days, with a data length of M data points. Construct prompts for the temperature time-series data sequence: Based on the historical operating patterns of the main steam temperature of this thermal power generating unit, analyze its long-term trend, diurnal seasonal fluctuations, and start-stop cycle characteristics under different loads, and generate physically logical temperature values ​​for the missing locations. The processing intent is to generate physically logical temperature values ​​for the missing locations. By inputting the prompts and the main steam temperature time-series data sequence into a large language model, the large language model combines the historical operating patterns of the main steam temperature to analyze its long-term trend, diurnal seasonal fluctuations, and start-stop cycle characteristics under different loads, generating filler values ​​for each missing point based on the fluctuation characteristics of normal data. For example, the missing temperature value during the unit's high-load operation period is filled with a value consistent with the steam temperature stabilizing at around 540℃ under high load. Furthermore, the data processing results output by the large language model can be manually verified to further improve the accuracy of data processing.

[0123] After obtaining the latest data processing results, they are input into the digital twin model and fault early warning model of the unit to update them. Through multiple updates, the prediction error of the digital twin model for the unit's main steam temperature and the accuracy of fault prediction can be improved, effectively avoiding false alarms and missed alarms caused by poor data quality. Therefore, the method provided in this application, while solving the problems of missing original data and noise, provides high-quality data input for the digital twin model and early warning model, significantly improving the model's prediction accuracy and achieving precise monitoring of the unit's operating status, thus providing strong data support for the safe and efficient operation of the equipment.

[0124] In summary, this application employs a large language model to deeply mine the long-term trends, seasonality, and periodicity of industrial time-series data during data processing. The missing values ​​filled in perfectly match the physical logic of equipment operation, avoiding subjective assumptions inherent in traditional methods. This ensures the integrity and continuity of the time series, resulting in high filling accuracy and strong data integrity. Furthermore, by learning historical data features, the large language model can accurately identify and correct complex interference noise, effectively improving the signal-to-noise ratio of the original data. This provides clean and reliable data input for subsequent models, guaranteeing the accuracy of subsequent model analysis and prediction. Moreover, by establishing a closed-loop mechanism for automatic new data processing and dynamic model updates, the early warning model and digital twin model can continuously learn new characteristics of equipment operation, effectively adapting to data changes over time and maintaining high-precision analysis and prediction performance over the long term. In other words, the method provided in this application achieves full automation from data input, feature extraction, preprocessing to model updates, eliminating the need for complex manual feature engineering and parameter tuning. This significantly reduces the manpower and time costs of industrial data preprocessing and has extremely high engineering application value.

[0125] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0126] Based on the same inventive concept, this application also provides a large-model-based data processing apparatus for implementing the large-model-based data processing method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more embodiments of the large-model-based data processing apparatus provided below can be found in the limitations of the large-model-based data processing method described above, and will not be repeated here.

[0127] In one exemplary embodiment, such as Figure 5 As shown, a data processing device based on a large model is provided, including: an M module, an N module, and an L module, wherein:

[0128] The acquisition module 501 is used to acquire equipment data collected from the production equipment and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder; the first processing module 502 is used to generate first sub-information matching the equipment data according to the data type of the equipment data, and fill the first sub-information into the first variable placeholder; the second processing module 503 is used to determine the processing intention for the equipment data, generate second sub-information matching the processing intention, and fill the second sub-information into the second variable placeholder; the third processing module 504 is used to use the prompt word template after filling in the information as prompt information, and guide the large language model to process the equipment data through the prompt information to obtain a data processing result matching the processing intention.

[0129] In one embodiment, the processing intent includes correcting abnormal data in the device data; the third processing module 504 is further configured to: obtain a preset data range corresponding to the production equipment under the data type; input the prompt information and the device data into the large language model to obtain an initial result matching the processing intent; the initial result includes sensor data corresponding to multiple times; if there is abnormal data in each of the sensor data that does not match the preset data range, correct the abnormal data based on the preset data range to obtain updated data; and obtain a data processing result matching the processing intent based on each of the updated data.

[0130] In one embodiment, the third processing module 504 is further configured to: obtain combined data by combining each of the updated data and normal data from each of the sensor data that match the preset data range; obtain feedback information for the combined data; and, if the feedback information characterization verification is passed, determine the combined data as a data processing result that matches the processing intention.

[0131] In one embodiment, the processing intent includes correcting abnormal data in the equipment data; the third processing module 504 is further configured to: acquire historical normal data and historical abnormal data of the production equipment under the data type; guide the large language model to identify and correct abnormal data in the equipment data with reference to the difference between the historical normal data and the historical abnormal data through the prompt information, and obtain a data processing result that matches the processing intent.

[0132] In one embodiment, the device data includes multiple sub-data corresponding to different times, and the processing intent includes determining the data location of abnormal data in the device data based on the data change amplitude between adjacent sub-data in the device data; the third processing module 504 is further configured to: guide the large language model through the prompt information, output the location of abnormal data in the device data based on the data change amplitude between adjacent sub-data in the device data; correct the sub-data corresponding to the abnormal data location, and obtain a data processing result matching the processing intent.

[0133] In one embodiment, the third processing module 504 is further configured to: determine the data change trend of the device data, generate third sub-information matching the data change trend, and fill the third sub-information into the prompt information; guide the large language model to correct the sub-data corresponding to the abnormal data position based on the data range represented by the common trend in the data change trend through the prompt information after filling the information, so as to obtain a data processing result matching the processing intention.

[0134] Each module in the aforementioned large-scale model-based data processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0135] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data from the large-scale data processing process. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a large-scale data processing method.

[0136] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a data processing method based on a large model. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0137] Those skilled in the art will understand that Figure 6 or Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0138] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0139] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0140] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0141] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0142] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0143] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0144] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A data processing method based on a large model, characterized in that, The method includes: Acquire equipment data collected for the production equipment, and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder; Based on the data type of the device data, generate first sub-information that matches the device data, and fill the first sub-information into the first variable placeholder; Determine the processing intent for the device data, generate second sub-information matching the processing intent, and fill the second sub-information into the second variable placeholder; The prompt word template after filling in the information is used as prompt information, and the prompt information guides the large language model to process the device data to obtain a data processing result that matches the processing intention.

2. The method according to claim 1, characterized in that, The processing intent includes correcting abnormal data in the device data; the step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes: Obtain the preset data range corresponding to the data type of the production equipment; The prompt information and the device data are input into the large language model to obtain an initial result that matches the processing intent; the initial result includes sensor data corresponding to multiple time points. If there is abnormal data in the sensor data that does not match the preset data range, the abnormal data is corrected based on the preset data range to obtain updated data; Based on the updated data, a data processing result matching the processing intent is obtained.

3. The method according to claim 2, characterized in that, The step of obtaining a data processing result matching the processing intent based on each of the updated data includes: Combined data is obtained by combining the updated data and the normal data from each of the sensor data that matches the preset data range; Obtain feedback information regarding the combined data; If the feedback information verification is successful, the combined data will be determined as the data processing result that matches the processing intention.

4. The method according to claim 1, characterized in that, The processing intent includes correcting abnormal data in the device data; The step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes: Obtain historical normal data and historical abnormal data of the production equipment under the data type described above; The prompts guide the large language model to identify and correct abnormal data in the device data by referring to the differences between the historical normal data and the historical abnormal data, thereby obtaining a data processing result that matches the processing intent.

5. The method according to claim 1, characterized in that, The device data includes multiple sub-data corresponding to different times, and the processing intent includes determining the data location of abnormal data in the device data based on the data change amplitude between adjacent time sub-data in the device data. The step of guiding the large language model to process the device data through the prompt information to obtain a data processing result matching the processing intent includes: The prompt information guides the large language model to output the location of abnormal data in the device data based on the magnitude of data change between adjacent time-series sub-data in the device data. Correct the sub-data corresponding to the abnormal data location to obtain a data processing result that matches the processing intention.

6. The method according to claim 5, characterized in that, The step of correcting the sub-data corresponding to the location of the abnormal data to obtain a data processing result that matches the processing intent includes: Determine the data change trend of the device data, generate third sub-information that matches the data change trend, and fill the third sub-information into the prompt information; The prompt information after filling in the information guides the large language model to correct the sub-data corresponding to the abnormal data location based on the data range represented by the common trend in the data change trend, so as to obtain a data processing result that matches the processing intention.

7. A data processing device based on a large model, characterized in that, The device includes: The acquisition module is used to acquire equipment data collected from the production equipment and a prompt word template matching the production equipment; the prompt word template includes a first variable placeholder and a second variable placeholder; The first processing module is configured to generate first sub-information matching the device data according to the data type of the device data, and fill the first sub-information into the first variable placeholder; The second processing module is used to determine the processing intention for the device data, generate second sub-information that matches the processing intention, and fill the second sub-information into the second variable placeholder. The third processing module is used to take the prompt word template after filling in the information as prompt information, and guide the large language model to process the device data through the prompt information to obtain a data processing result that matches the processing intention.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images