Industrial Internet of Things data analysis platform based on edge computing and big data

Abstract

The invention discloses an industrial internet of things data analysis platform based on edge calculation and big data, and relates to the technical field of industrial data, the industrial internet of things data analysis platform comprises a big database, a processing platform, an edge calculation module and a data acquisition module, the big database establishes a plurality of production models based on historical data, and the processing platform is used for processing the data. Each production model is labeled and classified by the big database; the processing platform establishes a corresponding model demand based on a preset production plan, matches the model demand with each production model in the big data, and screens out a production model meeting the requirement; a plurality of data acquisition modules are arranged, and each data acquisition module acquires actual production data of corresponding production equipment; according to the method, the comprehensive scoring mechanism of the production model is introduced, the matching degree between the production model and the actual production requirement is dynamically evaluated, the weight, the sensitivity factor and the deviation amplitude of each field are comprehensively considered, the optimal model is selected, and the production process is dynamically adjusted.

Description

Technical Field

[0001] The present invention relates to the field of industrial data technology, and specifically to an industrial Internet of Things data analysis platform based on edge computing and big data. Background Art

[0002] With the widespread application of industrial Internet of Things technology, real-time collection, processing and analysis of production data have become key means for industrial enterprises to improve production efficiency and product quality. Traditional industrial monitoring systems mainly rely on centralized data processing methods, that is, uploading real-time data of each production equipment to the cloud or central processing platform for unified analysis. Although this method can comprehensively monitor the overall production situation, it has significant performance bottlenecks when facing large-scale equipment and complex production environments;

[0003] The introduction of edge computing technology provides a distributed data processing method for the Industrial Internet of Things. Edge nodes perform preliminary analysis of data at locations close to production equipment, reducing the computing pressure on the central platform.

[0004] After searching, a Chinese patent (publication number: CN114896052A) discloses an industrial edge computing platform, which includes an edge gateway module, an edge cache module, a process gateway module, an industrial AI module, and an integrated development tool module; wherein, the edge gateway module is used to collect edge-side data; the edge cache module is used to store edge data through a data acquisition engine; the process gateway module is used to pre-process the cached edge data to achieve data association; the industrial AI module is used to perform data analysis on edge data; and the integrated development tool module is used to provide one-stop data deployment management.

[0005] In the prior art, if all data are uploaded frequently, network bandwidth resources will be occupied by redundant data, which is more serious in scenarios with a large number of devices and a fast production pace. This will lead to sudden abnormal situations in the production process, and the abnormal data cannot be quickly prioritized. There is a lag in the discovery and response of problems. Therefore, the present invention proposes an industrial Internet of Things data analysis platform based on edge computing and big data. Summary of the invention

[0006] The purpose of the present invention is to provide an industrial Internet of Things data analysis platform based on edge computing and big data to solve the problems mentioned in the above background technology.

[0007] The present invention can be implemented by the following technical solutions: an industrial Internet of Things data analysis platform based on edge computing and big data, including a big database, a processing platform, an edge computing module and a data acquisition module;

[0008] The big database is based on historical data and has multiple production models, and the big database labels and classifies each production model;

[0009] The processing platform establishes corresponding model requirements based on a preset production plan, where the production plan includes production processes, resource requirements, production equipment, and production targets, and the processing platform converts the preset production plan into a standardized data format when establishing the corresponding model requirements;

[0010] And the processing platform matches each production model in the big data based on the model requirements. When matching, the processing platform compares the fields in the production plan data format with the annotations and classifications of each production model item by item to select the production model that meets the requirements;

[0011] The requirements that are met include:

[0012] a1. The fields in the production plan data format completely match the annotations and classifications of the corresponding production model;

[0013] a2. There are differences in the fields of the production plan data format and the annotation and classification of the corresponding production model, but the product quality meets the requirements and the difference is within the preset range;

[0014] There are multiple data acquisition modules, and each data acquisition module is matched with a group of production equipment and collects actual production data of the corresponding production equipment;

[0015] There are multiple edge computing modules, and each edge computing module covers a corresponding production area respectively, and the edge computing module selects a corresponding model part from the production model of the processing platform as a matching sub-model based on the generation area;

[0016] At the same time, the edge computing module receives the actual production data of each production equipment in the corresponding production area through the corresponding data acquisition module, and matches the actual production data with the corresponding matching sub-model. The edge computing module collects the data deviation between the actual production data and the matching sub-model through time series to obtain the deviation sequence, and uploads the deviation sequence to the processing platform;

[0017] After receiving the deviation sequence, the processing platform performs distribution fitting between the current deviation sequence and the historical production data, determines whether the current production status conforms to the historical law, and predicts the future production status through a machine learning algorithm, including:

[0018] Predict changes in production quality that may be caused by fluctuations. Based on the relationship between fluctuations and production quality in historical data, a regression model or classification model is trained to predict changes in quality that may be caused by current fluctuations.

[0019] Predict whether equipment may fail or deviate. It is based on time series prediction algorithms to identify trends and abnormal points in equipment parameters and provide early warning of possible failures.

[0020] A further technical improvement of the present invention is that when there are differences between the annotation and classification of the fields in the production plan data format and the corresponding production model, but the product quality meets the requirements and the difference is within a preset range, the step of the processing platform screening the production model includes:

[0021] A1. Screening of preliminary matching models:

[0022] Compare the fields of the production plan with the annotations of the production model, and select the production models that are similar to the production plan in terms of fields. Set a difference tolerance threshold to eliminate the production models that exceed the difference tolerance threshold, so as to eliminate the production models with too large differences.

[0023] A2. According to the size of the field differences, the initially screened production models are sorted from small to large, and a matching set is constructed;

[0024] A3. For each production model in the matching set, verify whether the final product quality in its historical data is consistent with the target quality of the production plan, and eliminate the production models whose product quality does not meet the requirements;

[0025] A4. Calculate the difference impact factor;

[0026] The impact factor is calculated by associating the difference value with the quality change in the historical data, establishing a quantitative relationship between the field and the product quality, and using the historical data to calculate the impact factor of each field through regression analysis and sensitivity analysis methods;

[0027] A6. According to the field differences and corresponding influencing factors of the selected model, adjust the model parameters to better meet the production plan requirements;

[0028] A7. In the actual production process of using the revised production model, real-time data of the S time period is continuously collected, and the real-time data is dynamically compared with the revised production model;

[0029] In the S time period, if the deviation between the actual data and the production model is within the preset deviation threshold, the revised production model is verified to be qualified;

[0030] The computing platform also stores new production data and model modification records in a large database, and updates historical data and impact factor calculation models to improve the accuracy of the next match;

[0031] During the S time period, if the deviation between the actual data and the production model exceeds the preset deviation threshold, the production model parameters are further adjusted.

[0032] A further technical improvement of the present invention is that in step A7, the step of further adjusting the production model parameters includes:

[0033] Z1. Collect actual production data, compare the actual production data with the production plan target data, and calculate the deviation value;

[0034] Z2. Use regression analysis tools to determine which fields’ deviations have the greatest impact on product quality;

[0035] Z3, through the formula P i Q =P i X -αΔP i , adjust the production model parameters corresponding to the fields determined in Z2;

[0036] Where P i Q is the parameter value of the i-th field after correction, α is the learning rate, which is used to control the adjustment amplitude and avoid over-adjustment;

[0037] When the deviation is large, increase the learning rate to converge quickly;

[0038] When the deviation is small, reduce the learning rate and make fine adjustments;

[0039] Z4. After adjusting the production model parameters, repeat step A7 until the deviation between the actual data and the production model is within the preset deviation threshold range within the S time period.

[0040] A further technical improvement of the present invention is that if deviations of multiple fields exist simultaneously in the production model, a multivariate optimization method is used to adjust multidimensional parameters simultaneously, and the optimization goal is:

[0041] In the formula, min means that the optimization goal is to minimize the weighted sum of squares of the total deviations; ω i is the weight of field i; ΔP i is the difference between the production plan and the production model in field i;

[0042] Through the optimization algorithm, a set of parameter values ​​is found to minimize the sum of squares of weighted deviations, thus reducing production problems caused by parameter deviations.

[0043] A further technical improvement of the present invention is that: when each of the data acquisition modules acquires actual production data, a locally generated timestamp is added to the actual production data, and the timestamp is calibrated by the edge computing module according to the unified time source of the processing platform;

[0044] The edge computing module performs preliminary verification, sorting and filtering of the data from multiple acquisition modules, generates time series production data according to the set sampling frequency, and matches the time series production data with the matching sub-model.

[0045] A further technical improvement of the present invention is that: the edge computing module is provided with a corresponding deviation threshold based on the production equipment in the corresponding production area;

[0046] When receiving actual production data, the edge computing module marks the corresponding production equipment as stable or unstable based on the degree of matching between the actual production data and the deviation threshold;

[0047] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data is within the preset deviation threshold range and the holding time exceeds the preset holding time, the edge computing module marks the corresponding production equipment as a stable state. For the actual production data generated by the production equipment in a stable state, the edge computing module performs data downsampling or compression, and uploads the data in a low-frequency and low-computing-power manner;

[0048] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data exceeds the preset deviation threshold range, the edge computing module will mark the corresponding production equipment as an unstable state. For the actual production data generated by the production equipment in an unstable state, the edge computing module will upload the complete actual production data in real time, and store the actual production data within the V time range before the time point when the actual production data exceeds the deviation threshold, to facilitate subsequent calls.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention introduces a comprehensive scoring mechanism for production models to dynamically evaluate the degree of match between production models and actual production needs. By comprehensively considering the weight, sensitivity factor and deviation amplitude of each field, the system can select the optimal model from multiple candidate models and dynamically adjust the production process, thus avoiding the limitations of static models, making the production model more suitable for the current actual production environment, and significantly improving production efficiency and product quality.

[0051] In addition, the present invention quantitatively analyzes the impact of each field deviation on product quality through sensitivity factors, and realizes accurate optimization of the production model based on the adjustment formula, which can quickly converge to the production target requirements, effectively reduce the adverse impact of parameter deviation on production results, and further improve the stability and consistency of product quality;

[0052] At the same time, the present invention introduces a deviation sequence monitoring mechanism in the edge computing module, and dynamically marks the operating status of the production equipment according to the deviation status between the actual production data and the matching model. When the deviation is within the threshold range and stable, the equipment is marked as "stable state", and the edge module only uploads low-frequency summary data; when the deviation exceeds the threshold, the equipment is marked as "unstable state", and the edge module uploads complete data and deviation sequence to the processing platform in real time. This hierarchical management method not only reduces the communication and computing burden in the stable state, but also ensures a rapid response to abnormal conditions, improving the overall operating efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0054] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0055] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0056] Example 1

[0057] An industrial Internet of Things data analysis platform based on edge computing and big data, including a big database, a processing platform, an edge computing module and a data acquisition module;

[0058] The big database is based on historical data and has established multiple production models. The big database also labels and classifies each production model. The labeling and classification include:

[0059] Production goals: including output goals, quality requirements, and energy consumption goals;

[0060] Resource requirements: including types of raw materials and equipment models;

[0061] Process conditions: including key parameters (such as temperature, pressure, speed) and operating procedures;

[0062] Time and environment: including production cycle and environmental conditions required for the process;

[0063] Product quality: final performance data of the product in historical production;

[0064] The processing platform establishes corresponding model requirements based on the preset production plan. The production plan includes production process, resource requirements, production equipment, and production goals. When establishing the corresponding model requirements, the processing platform converts the preset production plan into a standardized data format, including the following fields:

[0065] 1) Basic information: production plan ID, creation time, urgency;

[0066] 2) Production requirements:

[0067] Target output: quantity, accuracy, and pass rate.

[0068] Resource allocation: raw material type, equipment model, and personnel requirements.

[0069] Process parameters: key production conditions that need to be met (such as temperature and pressure).

[0070] Time arrangement: plan start and end time, stage nodes;

[0071] And the processing platform matches each production model in the big data based on the model requirements. When matching, the processing platform compares the fields in the production plan data format with the annotations and classifications of each production model item by item to select the production model that meets the requirements;

[0072] The requirements that meet the requirements are:

[0073] The fields of the production plan data format fully match the annotations and classifications of the corresponding production model;

[0074] There are multiple data acquisition modules, and each data acquisition module is matched with a group of production equipment and collects actual production data of the corresponding production equipment;

[0075] There are multiple edge computing modules, and each edge computing module covers a corresponding production area respectively, and the edge computing module selects a corresponding model part from the production model of the processing platform as a matching sub-model based on the generation area;

[0076] At the same time, the edge computing module receives the actual production data of each production equipment in the corresponding production area through the corresponding data acquisition module. When each data acquisition module obtains the actual production data, it adds a locally generated timestamp to the actual production data. The timestamp is calibrated by the edge computing module according to the unified time source of the processing platform, and the time series production data is generated according to the set sampling frequency.

[0077] And after the edge computing module preliminarily verifies, sorts and filters the time series production data of multiple acquisition modules, it matches the time series production data with the corresponding matching sub-model to obtain the deviation sequence, and uploads the deviation sequence to the processing platform;

[0078] After receiving the deviation sequence, the processing platform performs distribution fitting between the current deviation sequence and the historical production data to determine whether the current production status conforms to the historical law, and predicts the future production status through machine learning algorithms, including:

[0079] Predict changes in production quality that may be caused by fluctuations. Based on the relationship between fluctuations and production quality in historical data, a regression model or classification model is trained to predict changes in quality that may be caused by current fluctuations.

[0080] Predict whether equipment may fail or deviate. It is based on time series prediction algorithms to identify trends and abnormal points in equipment parameters and provide early warning of possible failures.

[0081] The edge computing module is set with corresponding deviation thresholds based on the production equipment in the corresponding production area;

[0082] When receiving actual production data, the edge computing module marks the corresponding production equipment as stable or unstable based on the degree of match between the actual production data and the deviation threshold;

[0083] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data is within the preset deviation threshold range and the holding time exceeds the preset holding time, the edge computing module marks the corresponding production equipment as a stable state. For the actual production data generated by the production equipment in a stable state, the edge computing module performs data downsampling or compression, and uploads the data in a low-frequency and low-computing-power manner;

[0084] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data exceeds the preset deviation threshold range, the edge computing module will mark the corresponding production equipment as an unstable state. For the actual production data generated by the production equipment in an unstable state, the edge computing module will upload the complete actual production data in real time, and store the actual production data within the V time range before the time point when the actual production data exceeds the deviation threshold, to facilitate subsequent calls.

[0085] Example 2

[0086] An industrial Internet of Things data analysis platform based on edge computing and big data, including a big database, a processing platform, an edge computing module and a data acquisition module;

[0087] The big database is based on historical data and has established multiple production models. The big database also labels and classifies each production model.

[0088] The processing platform establishes corresponding model requirements based on the preset production plan;

[0089] And the processing platform matches each production model in the big data based on the model requirements. When matching, the processing platform compares the fields in the production plan data format with the annotations and classifications of each production model item by item to select the production model that meets the requirements;

[0090] The requirements that meet the requirements are:

[0091] The fields in the production plan data format are different from the annotations and classifications of the corresponding production model, but the product quality meets the requirements and the difference is within the preset range;

[0092] There are multiple data acquisition modules, and each data acquisition module is matched with a group of production equipment and collects actual production data of the corresponding production equipment;

[0093] There are multiple edge computing modules, and each edge computing module covers a corresponding production area respectively, and the edge computing module selects a corresponding model part from the production model of the processing platform as a matching sub-model based on the generation area;

[0094] At the same time, the edge computing module receives the actual production data of each production equipment in the corresponding production area through the corresponding data acquisition module. When each data acquisition module obtains the actual production data, it adds a locally generated timestamp to the actual production data. The timestamp is calibrated by the edge computing module according to the unified time source of the processing platform, and the time series production data is generated according to the set sampling frequency.

[0095] And after the edge computing module preliminarily verifies, sorts and filters the time series production data of multiple acquisition modules, it matches the time series production data with the corresponding matching sub-model to obtain the deviation sequence, and uploads the deviation sequence to the processing platform;

[0096] After receiving the deviation sequence, the processing platform performs distribution fitting between the current deviation sequence and the historical production data to determine whether the current production status conforms to the historical law, and predicts the future production status through machine learning algorithms;

[0097] When there are differences between the annotation and classification of the fields in the production plan data format and the corresponding production model, but the product quality meets the requirements and the difference is within the preset range, the processing platform screens the production model, including:

[0098] A1. Screening of preliminary matching models:

[0099] Compare the fields of the production plan with the annotations of the production model, and select the production models that are similar to the production plan in terms of fields. Set a difference tolerance threshold to eliminate the production models that exceed the difference tolerance threshold, so as to eliminate the production models with too large differences.

[0100] A2. According to the size of the field differences, the initially screened production models are sorted from small to large, and a matching set is constructed;

[0101] A3. For each production model in the matching set, verify whether the final product quality in its historical data is consistent with the target quality of the production plan, and eliminate the production models whose product quality does not meet the requirements;

[0102] A4. Calculate the difference impact factor;

[0103] Defining Impact Factor I i is the degree of influence of a field difference on the quality of the final product, and the formula is:

[0104] In the formula, ΔQ is the change in product quality, ΔP i is the difference between the production plan and the production model in field i;

[0105] A5. For each production model in the matching set, the comprehensive score is calculated by combining the field difference and the impact factor:

[0106] In the formula, S j is the comprehensive score of production model j; ω i is the weight of field i; the lower the comprehensive score, the more the production model conforms to the production plan;

[0107] A6. According to the field differences and corresponding influencing factors of the selected model, adjust the model parameters to better meet the production plan requirements. The adjustment formula is: P i X =P i m +ΔP i I i ;

[0108] Where P i X is the parameter value of the modified production model on the i-th field. The goal is to make the parameters of the production model as close as possible to the corresponding field value in the production plan; P i m The original parameter value of the currently selected production model in the i-th field is the original setting value of the production model in the database;

[0109] A7. In the actual production process of using the revised production model, real-time data of the S time period is continuously collected, and the real-time data is dynamically compared with the revised production model;

[0110] In the S time period, if the deviation between the actual data and the production model is within the preset deviation threshold, the revised production model is verified to be qualified;

[0111] The computing platform also stores new production data and model modification records in a large database, and updates historical data and impact factor calculation models to improve the accuracy of the next match;

[0112] During the S time period, if the deviation between the actual data and the production model exceeds the preset deviation threshold, the production model parameters are further adjusted, including:

[0113] Z1. Collect actual production data, compare the actual production data with the production plan target data, and calculate the deviation value;

[0114] Z2. Use regression analysis tools to determine which fields’ deviations have the greatest impact on product quality;

[0115] Z3, through the formula P i Q =P i X -αΔP i , adjust the production model parameters corresponding to the fields determined in Z2;

[0116] Where P i Q is the parameter value of the i-th field after correction, α is the learning rate, which is used to control the adjustment amplitude and avoid over-adjustment;

[0117] When the deviation is large, increase the learning rate to converge quickly;

[0118] When the deviation is small, reduce the learning rate and make fine adjustments;

[0119] Z4. After adjusting the production model parameters, repeat step A7 until the deviation between the actual data and the production model is within the preset deviation threshold range within the S time period;

[0120] The edge computing module is set with corresponding deviation thresholds based on the production equipment in the corresponding production area;

[0121] When receiving actual production data, the edge computing module marks the corresponding production equipment as stable or unstable based on the degree of match between the actual production data and the deviation threshold;

[0122] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data is within the preset deviation threshold range and the holding time exceeds the preset holding time, the edge computing module marks the corresponding production equipment as a stable state. For the actual production data generated by the production equipment in a stable state, the edge computing module performs data downsampling or compression, and uploads the data in a low-frequency and low-computing-power manner;

[0123] When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data exceeds the preset deviation threshold range, the edge computing module will mark the corresponding production equipment as an unstable state. For the actual production data generated by the production equipment in an unstable state, the edge computing module will upload the complete actual production data in real time, and store the actual production data within the V time range before the time point when the actual production data exceeds the deviation threshold, to facilitate subsequent calls.

[0124] Example 3

[0125] Compared with Example 2, the step of screening the production model by the processing platform in Example 3 includes:

[0126] A1. Screening of preliminary matching models:

[0127] Compare the fields of the production plan with the annotations of the production model, and select the production models that are similar to the production plan in terms of fields. Set a difference tolerance threshold to eliminate the production models that exceed the difference tolerance threshold, so as to eliminate the production models with too large differences.

[0128] A2. According to the size of the field differences, the initially screened production models are sorted from small to large, and a matching set is constructed;

[0129] A3. For each production model in the matching set, verify whether the final product quality in its historical data is consistent with the target quality of the production plan, and eliminate the production models whose product quality does not meet the requirements;

[0130] A4. Calculate the difference impact factor;

[0131] Defining Impact Factor I i is the degree of influence of a field difference on the quality of the final product, and the formula is:

[0132] In the formula, ΔQ is the change in product quality, ΔP i is the difference between the production plan and the production model in field i;

[0133] If deviations of multiple fields exist simultaneously in the production model, a multivariate optimization method is used to adjust multidimensional parameters simultaneously, and the optimization goal is:

[0134] In the formula, min means that the optimization goal is to minimize the weighted sum of squares of the total deviations; ω i is the weight of field i; ΔP i is the difference between the production plan and the production model in field i;

[0135] Through the optimization algorithm, a set of parameter values ​​is found to minimize the sum of squares of weighted deviations, thus reducing production problems caused by parameter deviations;

[0136] And the impact factor I i The calculation of introduces multivariate regression analysis:

[0137]

[0138] In the formula, And β 0 It represents the intercept term, which reflects the basic impact value on product quality change when all parameter deviations are zero; is the core item, representing the difference value ΔP i Linear contribution to product quality variation; β i is the regression coefficient, indicating the difference value ΔP i The weight of the product quality change value; ε is the error term;

[0139] P i is the parameter of the i-th field, is the time derivative of the i-th field parameter, and k is the adjustment factor;

[0140] A5. For each production model in the matching set, the comprehensive score is calculated by combining the field difference and the impact factor:

[0141] In the formula, S j is the comprehensive score of production model j; ω i is the weight of field i; the lower the comprehensive score, the more the production model conforms to the production plan;

[0142] A6. According to the field differences and corresponding influencing factors of the selected model, adjust the model parameters to better meet the production plan requirements. The adjustment formula is: P i X =P i m +ΔP i I i ;

[0143] Where P i X is the parameter value of the modified production model on the i-th field. The goal is to make the parameters of the production model as close as possible to the corresponding field value in the production plan; P i m The original parameter value of the currently selected production model in the i-th field is the original setting value of the production model in the database;

[0144] A7. In the actual production process of using the revised production model, real-time data of the S time period is continuously collected, and the real-time data is dynamically compared with the revised production model;

[0145] In the S time period, if the deviation between the actual data and the production model is within the preset deviation threshold, the revised production model is verified to be qualified;

[0146] The computing platform also stores new production data and model modification records in a large database, and updates historical data and impact factor calculation models to improve the accuracy of the next match;

[0147] During the S time period, if the deviation between the actual data and the production model exceeds the preset deviation threshold, the production model parameters are further adjusted.

[0148] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An industrial Internet of Things data analysis platform based on edge computing and big data, including a big database, a processing platform, an edge computing module and a data acquisition module, characterized in that: The big database is based on historical data and has multiple production models, and the big database labels and classifies each production model; The processing platform establishes corresponding model requirements based on the preset production plan, and the processing platform matches each production model in the big data based on the model requirements to screen out production models that meet the requirements; There are multiple data acquisition modules, and each data acquisition module is matched with a group of production equipment and collects actual production data of the corresponding production equipment; There are multiple edge computing modules, and each edge computing module covers a corresponding production area respectively, and the edge computing module selects a corresponding model part from the production model of the processing platform as a matching sub-model based on the generation area; At the same time, the edge computing module receives the actual production data of each production equipment in the corresponding production area through the corresponding data acquisition module, and matches the actual production data with the corresponding matching sub-model. The edge computing module collects the data deviation between the actual production data and the matching sub-model through time series to obtain the deviation sequence, and uploads the deviation sequence to the processing platform; After receiving the deviation sequence, the processing platform performs distribution fitting on the current deviation sequence and the historical production data, determines whether the current production status conforms to the historical rules, and predicts the future production status through a machine learning algorithm.

2. According to the industrial Internet of Things data analysis platform based on edge computing and big data according to claim 1, it is characterized in that: Eligible production models include: a1. The fields in the production plan data format completely match the annotations and classifications of the corresponding production model; a2. There are differences in the marking and classification of the fields in the production plan data format and the corresponding production model, but the product quality meets the requirements and the difference is within the preset range.

3. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 2 is characterized in that: In a2, the steps of processing the platform to screen the production model include: A1. Screening of preliminary matching models: Compare the fields of the production plan with the annotations of the production model, and select the production model that has similar fields with the production plan; A2. According to the size of the field differences, the initially screened production models are sorted from small to large, and a matching set is constructed; A3. For each production model in the matching set, verify whether the final product quality in its historical data is consistent with the target quality of the production plan, and eliminate the production models whose product quality does not meet the requirements; A4. Calculate the impact factor of each field; A5. For each production model in the matching set, a comprehensive score is calculated by combining the field differences and impact factors; A6. According to the field differences and corresponding influencing factors of the selected model, adjust the model parameters to better meet the production plan requirements; A7. In the actual production process of using the revised production model, real-time data of the S time period is continuously collected, and the real-time data is dynamically compared with the revised production model; During the S time period, if the deviation between the actual data and the production model is within the preset deviation threshold, the revised production model is verified to be qualified.

4. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 3 is characterized in that: In step A7, within the S time period, if the deviation between the actual data and the production model exceeds the preset deviation threshold, the production model parameters are further adjusted, including: Z1. Collect actual production data, compare the actual production data with the production plan target data, and calculate the deviation value; Z2. Use regression analysis tools to determine which fields’ deviations have the greatest impact on product quality; Z3, through the formula P i Q =P i X -αΔP i , adjust the production model parameters corresponding to the fields determined in Z2; Where P i Q is the parameter value of the ith field after correction, α is the learning rate; Z4. After adjusting the production model parameters, repeat step A7 until the deviation between the actual data and the production model is within the preset deviation threshold range within the S time period.

5. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 1, characterized in that: If deviations of multiple fields exist simultaneously in the production model, a multivariate optimization method is used to adjust multidimensional parameters simultaneously, and the optimization goal is: In the formula, min means that the optimization goal is to minimize the weighted sum of squares of the total deviations; ω i is the weight of field i; ΔP i It is the difference value between the production plan and the production model in field i.

6. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 1, characterized in that: When each of the data acquisition modules acquires actual production data, a locally generated timestamp is added to the actual production data, and the timestamp is calibrated by the edge computing module according to the unified time source of the processing platform; The edge computing module performs preliminary verification, sorting and filtering of the data from multiple acquisition modules, generates time series production data according to the set sampling frequency, and matches the time series production data with the matching sub-model.

7. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 1 is characterized in that: The edge computing module is provided with a corresponding deviation threshold based on the production equipment in the corresponding production area; When receiving actual production data, the edge computing module marks the corresponding production equipment as a stable state or an unstable state based on the degree of matching between the actual production data and the deviation threshold.

8. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 7 is characterized in that: When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data is within the preset deviation threshold range and the holding time exceeds the preset holding time, the edge computing module marks the corresponding production equipment as a stable state. For the actual production data generated by the production equipment in a stable state, the edge computing module downsamples or compresses the data and uploads the data in a low-frequency and low-computing-power manner.

9. The industrial Internet of Things data analysis platform based on edge computing and big data according to claim 7 is characterized in that: When the deviation between the actual production data of the corresponding production equipment and the corresponding matching sub-model data exceeds the preset deviation threshold range, the edge computing module will mark the corresponding production equipment as an unstable state. For the actual production data generated by the production equipment in an unstable state, the edge computing module will upload the complete actual production data in real time, and store the actual production data within the V time range before the time point when the actual production data exceeds the deviation threshold.

Images