Real-time production process tracing monitoring method and system driven by Internet of Things

Abstract

The invention discloses a real-time production process tracing monitoring method and system driven by the Internet of Things, and relates to the technical field of manufacturing process management, and the method comprises the steps: monitoring a production line in real time according to the Internet of Things, and obtaining each production line monitoring block; detecting the product abnormity of each production line, and obtaining a product abnormity feature set of each production line; tracing from equipment, environment and other factors and constructing a production line abnormity tracing atlas; and finally, carrying out production line abnormity management according to the three types of maps to realize production process tracing and abnormity management and control. According to the invention, the technical problem that the production abnormality management is not comprehensive and not timely due to the fact that the traditional method is difficult to accurately trace multiple factors such as equipment and environment for the product abnormality of multiple production lines of a production workshop in the manufacturing process management is solved, and the purposes of accurately positioning the multiple factors such as the equipment and environment of the product abnormality and improving the production abnormality management efficiency are achieved. And the technical effects of comprehensiveness and timeliness of production abnormity management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of manufacturing process management, and in particular to a real-time production process traceability monitoring method and system driven by the Internet of Things. Background Art

[0002] In the field of manufacturing process management, real-time monitoring of the production process and accurate tracing of anomalies are crucial for ensuring product quality and production efficiency. Existing technologies often rely on manual inspections or single-dimensional data collection (e.g., focusing solely on equipment operating parameters), using fixed thresholds to detect product anomalies and trace equipment failures. These methods have limited applicability in closed production lines with stable production environments and single process parameters. However, as intelligent manufacturing demands greater transparency and refined management of production processes, their limitations are becoming increasingly apparent. Summary of the Invention

[0003] This application provides an IoT-driven real-time production process traceability monitoring method and system to address the technical problem in manufacturing process management where traditional methods have difficulty accurately tracing multiple factors such as equipment and environment for product anomalies on multiple production lines in a production workshop, resulting in incomplete and untimely production anomaly management.

[0004] In a first aspect, the present application provides a real-time production process traceability monitoring method driven by the Internet of Things, the method comprising: performing real-time monitoring on each production line of an Internet of Things production workshop to obtain monitoring blocks for each production line; performing product anomaly detection on each production line according to the monitoring blocks for each production line to obtain an abnormality feature set of each production line product; tracing the abnormality feature set of each production line product for equipment factors according to the monitoring blocks for each production line product to construct a first production line anomaly tracing map; tracing the abnormality feature set of each production line product for environmental factors according to the monitoring blocks for each production line product to construct a second production line anomaly tracing map; tracing the abnormality feature set of each production line product for other factors according to the monitoring blocks for each production line product to construct a third production line anomaly tracing map; and performing production line anomaly management according to the first production line anomaly tracing map, the second production line anomaly tracing map, and the third production line anomaly tracing map.

[0005] According to a second aspect of the present application, a real-time production process traceability monitoring system driven by the Internet of Things is provided, the system comprising: a production line monitoring block acquisition module for performing real-time monitoring on each production line of an Internet of Things production workshop to obtain each production line monitoring block; an abnormal feature set acquisition module for performing product abnormality detection on each production line according to each production line monitoring block to obtain an abnormal feature set of each production line product; a first map construction module for tracing the abnormal feature set of each production line product for equipment factors according to each production line monitoring block to construct a first map for production line abnormality tracing; a second map construction module for tracing the abnormal feature set of each production line product for environmental factors according to each production line monitoring block to construct a second map for production line abnormality tracing; a third map construction module for tracing the abnormal feature set of each production line product for other factors according to each production line monitoring block to construct a third map for production line abnormality tracing; and a production line abnormality management module for performing production line abnormality management according to the first, second and third production line abnormality tracing maps.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application relates to the field of manufacturing process management. By collecting multi-source data (equipment operating status, product characteristics, environmental parameters, etc.) from each production line monitoring block in the IoT production workshop, abnormal feature data is obtained through blockchain encapsulation, machine learning modeling, etc., and a multi-dimensional traceability map of equipment / environment / other factors is constructed. Combined with the dynamic matching of production control schemes and reinforcement learning causal analysis, the root cause of production process anomalies is accurately located, making the production anomaly detection and traceability results in manufacturing process management more comprehensive and reliable. Specifically, a dynamic anomaly detection space is constructed by real-time monitoring of production line data, key features are screened using a support-confidence model, and then multi-source data is associated through a reinforcement learning fusion causal model to form an interpretable anomaly causal network. This solves the problem of insufficient traceability accuracy of traditional methods in multi-factor coupling scenarios, provides technical support for intelligent and transparent management of the manufacturing process, and achieves the technical effect of accurately locating multiple factors such as equipment and environment that cause product anomalies, improving the comprehensiveness and timeliness of production anomaly management. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1This is a flow chart of the real-time production process traceability monitoring method driven by the Internet of Things provided in an embodiment of the present application.

[0009] Figure 2 This is a structural diagram of the real-time production process traceability monitoring system driven by the Internet of Things provided in an embodiment of the present application.

[0010] Explanation of the accompanying drawings: production line monitoring block acquisition module 1, abnormal feature set acquisition module 2, first map construction module 3, second map construction module 4, third map construction module 5, production line abnormality management module 6. DETAILED DESCRIPTION

[0011] This application provides an IoT-driven real-time production process traceability monitoring method and system to address the technical problem in manufacturing process management where traditional methods have difficulty accurately tracing multiple factors such as equipment and environment for product anomalies on multiple production lines in a production workshop, resulting in incomplete and untimely production anomaly management.

[0012] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0014] Example 1, as Figure 1 As shown, a real-time production process traceability monitoring method driven by the Internet of Things, wherein the method includes: Step A100: Real-time monitoring is performed on each production line in the IoT production workshop to obtain monitoring blocks for each production line.

[0015] In the embodiment of the present application, the IoT production workshop uses IoT technology to monitor each production line in real time, collect multi-source data such as equipment operation, product characteristics, environmental parameters, etc., and form a production line monitoring block through data cleaning and blockchain packaging, thereby realizing an intelligent production place for product anomaly detection, multi-factor traceability and anomaly management.

[0016] Specifically, based on the production line monitoring data set obtained from real-time monitoring of each production line in the IoT production workshop, the monitoring standard data set is obtained through data cleaning, and then the production line monitoring blocks are obtained through blockchain packaging processing. The specific steps are detailed in A110-A130.

[0017] Step A200: Perform product anomaly detection on each production line according to each production line monitoring block to obtain an abnormality feature set of each production line product.

[0018] Optionally, product feature data is identified according to each production line monitoring block and the production control plan of the corresponding production line is read. The product anomaly detection space of each production line is constructed through product feature mining, and anomaly detection is performed on the product feature data of each production line to generate anomaly feature sets of products of each production line. The specific steps are described in detail in A210-A240.

[0019] Step A300: tracing the abnormal feature set of each production line product based on the monitoring blocks of each production line to determine the equipment factors, and constructing a first graph for tracing the abnormality of the production line.

[0020] In one embodiment of the present application, the equipment monitoring data is identified based on each production line monitoring block and anomalies are identified. The association between each production line equipment and the product anomaly feature set is analyzed by tracing the cause, and the first production line anomaly tracing graph is generated after graphical sorting. The specific steps are described in detail in A310-A350.

[0021] Step A400: Tracing the environmental factors of the abnormal feature sets of the products of each production line according to the monitoring blocks of each production line, and constructing a second graph for tracing the abnormalities of the production line.

[0022] Specifically, environmental monitoring data is identified according to the monitoring blocks of each production line and anomalies are identified. The association between each production line environment and the product anomaly feature set is analyzed through tracing, and a second production line anomaly tracing graph is generated after graphical sorting. The specific steps are described in detail in A410-A440.

[0023] Step A500: Tracing other factors of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and constructing a third graph for tracing the abnormality of the production line.

[0024] Specifically, other monitoring data is identified and anomalies are identified based on the monitoring blocks of each production line. The association between other factors of each production line and the product abnormality feature set is analyzed through tracing, and the third production line anomaly tracing graph is generated after graphical sorting. The specific steps are detailed in A510-A540.

[0025] Step A600: Perform production line abnormality management according to the first production line abnormality tracing map, the second production line abnormality tracing map, and the third production line abnormality tracing map.

[0026] Specifically, the first, second, and third graphs are traced based on the production line anomalies to generate abnormality alarms for each production line. The specific steps are described in detail in A610.

[0027] Furthermore, step A200 in the method provided in the embodiment of the present application includes: A210: Perform product feature identification according to the production line monitoring blocks to obtain product feature data of the first production line, product feature data of the second production line, and product feature data of the Pth production line, where P is a positive integer.

[0028] A220: Read the production control plan for each of the production lines to obtain the production control plan for the first production line, the production control plan for the second production line, and so on, the production control plan for the Pth production line.

[0029] A230: Product feature mining is performed based on the first production line production control plan, the second production line production control plan...the Pth production line production control plan to construct a first product anomaly detection space, a second product anomaly detection space...the Pth product anomaly detection space.

[0030] A240: According to the first product anomaly detection space, the second product anomaly detection space...the Pth product anomaly detection space, anomaly detection is performed on the first production line product feature data, the second production line product feature data...the Pth production line product feature data respectively to generate the anomaly feature set of each production line product.

[0031] In the embodiments of this application, product features are information used to characterize product status. Production control plans are the current control parameters of each production line. The anomaly detection space is a model space constructed by mining product features based on the production control plans of each production line.

[0032] Specifically, product characteristics are first identified for each production line based on established monitoring blocks for each production line (trusted data encapsulated by blockchain). IoT sensors collect key characteristic data of products in real time during the production process. For example, physical parameters such as dimensional accuracy and surface roughness are collected for the first production line, while electrical performance indicators such as voltage and current fluctuations are collected for the second production line. This generates structured product characteristic data for the first production line, the second production line, and so on, and so on (where P is the total number of production lines), achieving a comprehensive digital representation of the product status of each production line.

[0033] Secondly, when reading the production control plan for each production line, the IoT interface retrieves the currently effective set of control parameters from the corresponding control parameter storage node (such as a database or edge computing device) in real time, forming a structured production control plan. For example, for production lines 1 through 1, the production control plans for lines 1 through 1 are read, including data such as equipment operating parameters (such as spindle speed and temperature thresholds), process standards (such as welding time and assembly accuracy), and quality specifications (such as dimensional tolerances and surface roughness requirements). A line control plan might specifically include real-time control parameters such as a speed of 1500 rpm for equipment A, a temperature control range of 20°C-25°C, and a dimensional tolerance of ±0.05 mm for product B. This provides a benchmark for subsequent product anomaly detection.

[0034] Next, based on the production control plan for each production line, a set of normal product feature records is retrieved. After support evaluation, confidence analysis, and confidence evaluation constraint selection, the first product anomaly detection space is constructed. The specific steps are detailed in A231-A234. Similarly, the second product anomaly detection space... and the Pth product anomaly detection space can be further constructed.

[0035] Finally, the real-time product feature data from each production line is input into the corresponding anomaly detection space for comparison and analysis. By calculating the degree of deviation between the real-time data and the normal features in the detection space, feature items that exceed the allowable range are identified, and an abnormal feature set for each production line's products is generated. Each abnormal feature set for each production line includes the first product abnormal feature set, the second product abnormal feature set, and so on. For example, if a dimensional parameter of a product on the first production line exceeds the ±0.1mm tolerance range, it is marked as an abnormal feature and included in the first product abnormal feature set.

[0036] Through a closed-loop process of real-time collection of production line data, precise reading of control schemes, dynamic construction of feature space, and intelligent identification of abnormal features, dynamic anomaly detection based on personalized production line standards is achieved, significantly improving the accuracy and comprehensiveness of product abnormal feature identification, and providing a reliable abnormal data foundation for subsequent multi-factor tracing.

[0037] Furthermore, step A230 in the method provided in the embodiment of the present application includes: A231: Perform normal product feature retrieval according to the first production line production control plan to obtain a product feature record set.

[0038] A232: Perform support evaluation based on the product feature record set to obtain a product feature support evaluation set.

[0039] A233: Confidence analysis is performed on the product feature record set according to the product feature support evaluation set to obtain a product feature confidence evaluation set.

[0040] A234: Based on the confidence evaluation constraint, the product feature record set is selected according to the product feature confidence evaluation set to obtain the first product anomaly detection space.

[0041] In this embodiment, support is a quantitative evaluation of the frequency of occurrence of each feature in the product feature record set, which is used to measure the universality of the feature. Confidence is the probability that the product is qualified when the feature appears, which is used to screen features that are strongly correlated with the product's qualified status.

[0042] Optionally, first, perform a normal product feature search based on the production control plan for the first production line. Extract the feature ranges of qualified products (such as dimensional tolerances and performance indicator thresholds) from the process standards and quality specifications included in the plan. Based on this range, retrieve product feature data that meets these standards from historical production data to form a product feature record set. For example, if a production line's control plan specifies that product length must be within the range of 100 ± 0.5 mm, retrieve the product features in the historical data that fall within this range to form a record set for that production line.

[0043] Next, we evaluate the support of the product feature record set. By counting the frequency of each feature in the record set (i.e., its support), we quantify its prevalence. For example, if the feature length 100mm±0.5mm appears 200 times in a record set with 250 records, its support is 80%. This forms the product feature support evaluation set, which is used to screen for frequently occurring core features.

[0044] Next, the record set is analyzed for confidence based on the support evaluation set. Confidence measures the strength of the association between a feature and the product's acceptable status and is calculated by calculating the probability of a product being acceptable when the feature appears. For example, if the product's acceptable rate is 95% when the feature surface roughness Ra ≤ 1.6 μm appears, the confidence level is 95%. This forms the product feature confidence evaluation set, eliminating features that are frequently encountered but weakly correlated with acceptable status (such as occasional, non-critical features).

[0045] Finally, based on pre-set confidence evaluation constraints (e.g., requiring a confidence level of ≥90%), features in the record set are selected. Only features that meet both support and confidence requirements are retained to construct the first product anomaly detection space. For example, a feature with a support level of 85% and a confidence level of 92% meets both thresholds and is included in the detection space as a normal feature benchmark.

[0046] Through the dual-threshold screening steps and the use of the support-confidence quantitative evaluation system, key features that are highly consistent with the production line process are dynamically screened out, and an anomaly detection space that is more in line with actual production requirements is constructed. This significantly improves the accuracy and reliability of product anomaly detection, and provides a scientific judgment basis for subsequent anomaly location based on trusted data.

[0047] Furthermore, step A300 in the method provided in the embodiment of the present application includes: A310: The abnormal feature set of each production line product includes the first product abnormal feature set, the second product abnormal feature set...the Pth product abnormal feature set.

[0048] A320: Identify equipment monitoring data according to each production line monitoring block to obtain a first production line equipment monitoring area, a second production line equipment monitoring area, and so on, a Pth production line equipment monitoring area.

[0049] A330: Identify abnormalities based on the first production line equipment monitoring area, the second production line equipment monitoring area...the Pth production line equipment monitoring area to obtain a first equipment abnormality feature set, a second equipment abnormality feature set...the Pth equipment abnormality feature set.

[0050] A340: Based on the first device abnormality feature set, the second device abnormality feature set...the Pth device abnormality feature set, the first product abnormality feature set, the second product abnormality feature set...the Pth product abnormality feature set are traced and analyzed respectively to obtain multiple product abnormality device traceability sets.

[0051] A350: Graphically combing the product abnormality feature sets of each production line and the multiple product abnormality equipment tracing cause sets to generate the first graph for tracing the production line abnormality.

[0052] In the embodiment of the present application, graphing refers to the structured sorting of the abnormal feature sets of each production line product and the abnormal feature sets of equipment / environment / other factors and their causal relationships, using nodes to represent abnormal features and directed edges to represent causal relationships to construct a visual graph.

[0053] Specifically, equipment monitoring data is first identified based on each production line monitoring block (encapsulated trusted data). Real-time equipment operating parameters, such as the vibration frequency, temperature, and rotational speed of equipment on the first production line, are extracted from the production line monitoring data stored on the blockchain. This creates an equipment monitoring zone for the first production line containing real-time data for multiple devices. Similarly, equipment monitoring zones for the second through P production lines are obtained. For example, a production line equipment monitoring zone might contain 10 operating parameters for three key pieces of equipment, reflecting their dynamic status in real time.

[0054] Next, we identify anomalies in the data from the equipment monitoring areas on each production line. Using preset equipment operating thresholds (e.g., vibration amplitude ≤ 50dB, temperature range 20°C-40°C), we detect parameter values outside the normal range and generate a first set of equipment anomaly signatures (e.g., abnormal vibration frequency of equipment A, temperature exceeding the limit of equipment B). Similarly, we generate anomaly signatures for equipment on other production lines.

[0055] Next, the device abnormality feature set and the product abnormality feature set are traced and analyzed to form a product abnormality device traceability set. The specific steps are described in detail in A341-A346.

[0056] Finally, the product anomaly feature sets and equipment traceability sets for each production line are organized into a graph. The specific process is as follows: First, the product anomaly feature sets for each production line (e.g., the anomaly feature sets for the first through P products) and the product anomaly equipment traceability sets derived through traceability analysis are summarized (each product anomaly feature set corresponds to a corresponding equipment traceability set). Then, using product anomaly features (e.g., surface cracks, dimensional deviations) and equipment anomaly features (e.g., abnormal vibration frequency of equipment A, spindle speed fluctuations) as graph nodes, the causal relationships identified through traceability analysis (e.g., abnormal vibration frequency of equipment → surface cracks of product) are structured as directed edges, organized according to the production line dimension. For example, if the temperature exceeding the limit of equipment B identified in the equipment monitoring area of production line 1 is related to the insufficient fracture strength of the product in the product anomaly feature set for that production line, after confirming the causal relationship through traceability analysis, a directed edge is established in the graph from the node for the temperature exceeding the limit of equipment B to the node for the insufficient fracture strength of the product, and attributes such as the association confidence are annotated. The complex relationship between equipment and product anomalies is then converted into a visual map, clearly presenting the impact path of equipment factors on product quality, forming the first map for tracing production line anomalies, and providing an intuitive basis for quickly locating the root cause of production anomalies.

[0057] Through the above steps, the precise location of product anomalies caused by equipment factors is achieved. Compared with existing technologies, this solution significantly improves the efficiency and accuracy of equipment traceability through trusted data support and intelligent traceability models, providing a visual decision-making basis for rapid response to production anomalies and preventive maintenance of equipment.

[0058] Furthermore, step A340 in the method provided in the embodiment of the present application includes: A341: Extract the Kth feature of the product anomaly based on the first product anomaly feature set, where K is a positive integer.

[0059] A342: Extract associated data from the first device abnormality feature set based on the Kth feature of the product abnormality to obtain the Kth associated device abnormality feature.

[0060] A343: Perform supervised training based on the product abnormal equipment tracing record set to obtain multiple product abnormal equipment tracing models.

[0061] A344: Establish a first tracing node model and a second tracing node model based on the multiple product abnormality equipment tracing models.

[0062] A345: Perform reinforcement learning fusion on the second tracing node model according to the first tracing node model to obtain a product abnormality equipment tracing channel.

[0063] A346: Input the Kth feature of the product abnormality and the Kth associated device abnormality feature into the product abnormal device tracing channel to obtain the Kth product abnormal device tracing result, and add the Kth product abnormal device tracing result to the first product abnormal device tracing set.

[0064] Specifically, first, the Kth feature of the product abnormality is extracted from the first product abnormality feature set (K is a positive integer). For example, after extracting the Kth feature of the product abnormality (such as surface cracks) from the first product abnormality feature set, the first equipment abnormality feature set is data mined through the association rule algorithm to calculate the support and confidence of the equipment operating parameters (such as vibration frequency, temperature, etc.) and the product abnormality feature, and screen out the equipment abnormality parameters with a correlation higher than a preset threshold, thereby obtaining the Kth associated equipment abnormality feature (such as abnormal vibration frequency of equipment A), and realizing the association analysis between product abnormality and equipment operating status.

[0065] Secondly, when using the random forest algorithm to establish a product abnormality equipment attribution model, the input is a historically accumulated set of product abnormality equipment attribution records. This record set contains verified equipment anomaly-product anomaly causal pairs, specifically covering equipment anomaly characteristics (such as abnormal parameters such as vibration frequency, temperature, and speed) and corresponding product anomaly characteristics (such as surface cracks and dimensional deviations). Feature engineering processing (such as normalization and discretization) is required to adapt the data to the algorithm input requirements. The algorithm proceeds as follows: multiple subsets are randomly selected from the original data through bootstrap sampling. A decision tree is constructed for each subset. When splitting a decision tree node, some features are randomly selected to calculate the Gini index or information gain to determine the splitting rule. The final result is a forest model composed of multiple decision trees. The output is multiple product abnormality equipment attribution models (base models). By learning the feature mapping relationships in the historical data, these models have the ability to predict the impact of equipment anomalies on product quality.

[0066] Next, these foundational models serve as the first-node models (base models) for attribution analysis, enabling them to make preliminary predictions about the association between a single device anomaly and a product anomaly, outputting preliminary causal association probabilities or conclusions. Next, a second-node model (meta-model) is constructed, integrating the predictions of multiple base models through ensemble learning methods (such as voting and weighted averaging). For complex scenarios where multiple device anomalies synergistically impact product quality, the meta-model's comprehensive reasoning improves the accuracy of causal relationship judgments. This hierarchical modeling process progresses from single-feature correlation analysis to comprehensive reasoning for complex scenarios, laying the foundation for the subsequent generation of reliable attribution channels for product anomaly analysis.

[0067] The two node models are then fused and optimized using a reinforcement learning algorithm. Initial predictions from the base model (such as the probability of association between a single device anomaly and a product anomaly) are used as input to the meta-model. A reinforcement learning reward function is then designed, assigning positive rewards to accurate attribution results (such as a verified causal relationship between a device anomaly and a product anomaly) and negative rewards to incorrect predictions or results with weak associations. Through continuous iterative training, the reward mechanism dynamically adjusts the meta-model's parameter weights, strengthening high-confidence association rules and weakening low-confidence rules. Ultimately, an adaptive product anomaly device attribution channel is formed. This channel dynamically optimizes association rules based on real-time input of device and product anomaly characteristics, accurately inferring the causal relationship between device and product anomalies.

[0068] Finally, the extracted Kth feature of the product anomaly and the associated equipment anomaly features are input into the product anomaly equipment tracing channel. The model outputs the Kth product anomaly equipment tracing result (e.g., abnormal vibration frequency of equipment A is the primary cause of product surface cracks). This result is added to the first product anomaly equipment tracing set, completing the tracing process for a single anomaly. Repeat these steps to obtain multiple product anomaly equipment tracing sets from the first to the Pth.

[0069] Through multi-model training and intelligent causal reasoning, the intelligent deduction of the causal relationship between equipment anomalies and product anomalies is achieved, significantly improving the efficiency and accuracy of root cause location of anomalies in complex production environments.

[0070] Furthermore, step A400 in the method provided in the embodiment of the present application includes: A410: Identify environmental monitoring data according to the production line monitoring blocks to obtain the first production line environmental monitoring area, the second production line environmental monitoring area...the Pth production line environmental monitoring area.

[0071] A420: Identify abnormalities based on the first production line environmental monitoring area, the second production line environmental monitoring area...the Pth production line environmental monitoring area to obtain a first environmental abnormality feature set, a second environmental abnormality feature set...the Pth environmental abnormality feature set.

[0072] A430: Based on the first environmental abnormality feature set, the second environmental abnormality feature set...the Pth environmental abnormality feature set, the first product abnormality feature set, the second product abnormality feature set...the Pth product abnormality feature set are traced and analyzed respectively to obtain multiple product abnormality environmental traceability sets.

[0073] A440: Graphically comb the abnormal feature sets of the products of each production line and the abnormal environmental cause sets of the multiple products to generate a second graph for tracing the abnormality of the production line.

[0074] Specifically, environmental monitoring data is first identified based on trusted data from each production line's monitoring area. IoT terminals deployed in the production workshop, such as temperature and humidity sensors and air quality monitoring equipment, collect real-time environmental parameters from each production line (such as the temperature, humidity, and dust particle concentration of the first production line). This generates structured data for the first production line's environmental monitoring area. Similarly, data for the second through P production lines' environmental monitoring areas is obtained. For example, the environmental monitoring area of a production line contains real-time data such as temperature (20°C-30°C), humidity (40%-60%RH), and dust concentration (≤1000ppm), covering key indicators of the production environment.

[0075] Next, anomalies are identified in the data from each production line's environmental monitoring area. Technical personnel compare real-time data with the standard range using the actual production line's preset environmental standard thresholds (e.g., a process temperature requirement of 25°C ± 2°C). They detect environmental parameters that exceed the allowable range and generate a first set of environmental anomaly signatures (e.g., temperature exceeding the limit, abnormally high humidity). Similarly, anomaly signatures for other production lines are generated. For example, if the real-time temperature on a production line is 35°C, exceeding the standard upper limit by 30°C, this is marked as a temperature anomaly signature.

[0076] Next, we perform causal analysis on the environmental anomaly feature set and the product anomaly feature set. We perform correlation analysis on the first environmental anomaly feature set (e.g., temperature exceeding the limit) and the first product anomaly feature set (e.g., blistering on the product surface), and train a causal model using historical data. The model building and training process is as follows: The input is a causal pair consisting of historically accumulated environmental monitoring data and product anomaly data (such as the corresponding records of abnormal environmental parameters such as temperature and humidity and abnormalities such as blistering and rust on the product surface). The data is trained through regression analysis and association rule algorithms (such as Apriori). Regression analysis is used to model the numerical association between environmental parameters and product anomalies, and association rule algorithms are used to mine the frequent co-occurrence patterns of environmental anomaly characteristics and product anomaly characteristics.

[0077] First, preprocess the environmental monitoring data (continuous parameters such as temperature and humidity) and product anomaly data from historical production data. By normalizing the environmental parameter values and discretizing the product anomaly characteristics, a structured dataset is formed. For regression analysis, a linear or nonlinear regression model (such as polynomial regression) is selected, using the environmental parameters as independent variables and quantitative indicators of product anomalies (such as blister rate and defect rate) as dependent variables. Model parameters are optimized using least squares or gradient descent methods to establish a numerical correlation model between environmental parameters and product anomalies. For example, a regression equation can be fitted that indicates a 30% increase in product surface blister rate for every 5°C increase in temperature. For the Apriori association rule algorithm, first set the support threshold (such as ≥80%) and confidence threshold (such as ≥90%), and perform frequent item set mining on the discretized environmental anomaly features (such as temperature exceeding the limit and humidity abnormality) and product anomaly features to generate association rules. For example, temperature exceeding the limit and humidity abnormality → product surface blistering (support 85%, confidence 92%). Cross-validation is used to ensure the reliability of the rules, and ultimately form a causal rule set for environmental anomalies and product anomalies for subsequent tracing and analysis.

[0078] The trained model can output the corresponding product abnormality features and causal relationship strength (such as temperature exceeding the limit → blistering on the product surface, with a confidence level of 95%) based on the input environmental abnormality feature set (such as temperature exceeding the limit), forming multiple product abnormality environment causal sets.

[0079] Finally, we graphically organize the product anomaly feature sets and environmental causal factors for each production line. (This step is similar to the previous one and will not be detailed here due to the limitations of this manual.) Using environmental and product anomaly features as nodes and causal relationships as directed edges, we construct a second graph for tracing production line anomaly. For example, the humidity anomaly node in the graph is connected to the product rust node via a directed edge, visually demonstrating the impact of environmental factors on product quality.

[0080] Through the process of real-time collection of environmental data, intelligent identification of abnormal characteristics, causal modeling analysis and graph visualization presentation, the lag and one-sidedness of environmental factor tracing in existing technologies are compensated, and the efficiency and accuracy of tracing the impact of environmental factors on product quality during the production process are significantly improved, providing a scientific basis for targeted adjustment of environmental control parameters and optimization of production processes.

[0081] Furthermore, step A500 in the method provided in the embodiment of the present application includes: A510: Identify other monitoring data based on the monitoring blocks of each production line to obtain other monitoring areas of the first production line, other monitoring areas of the second production line, and so on and so forth.

[0082] A520: Identify abnormalities based on the other monitoring areas of the first production line, the other monitoring areas of the second production line, ... the other monitoring areas of the P-th production line to obtain a first other abnormal feature set, a second other abnormal feature set, ... the P-th other abnormal feature set.

[0083] A530: Based on the first other abnormal feature set, the second other abnormal feature set...the Pth other abnormal feature set, the first product abnormal feature set, the second product abnormal feature set...the Pth product abnormal feature set are respectively traced and analyzed to obtain multiple product abnormality other traceability sets.

[0084] A540: Graphically comb the abnormal feature sets of the products of each production line and the other traceability sets of the multiple product abnormalities to generate a third graph for tracing the abnormalities of the production line.

[0085] In one embodiment, other monitoring data is first identified based on trusted data from each production line's monitoring area. The IoT collects multi-source, heterogeneous data beyond equipment and environmental data, such as raw material composition indicators (e.g., carbon content of a batch of steel), process parameter adjustment records (e.g., temporary extension of welding time), and operator authority change logs. This generates structured data from other monitoring areas on the first production line. Similarly, data from other monitoring areas on the second through P production lines is obtained. This data covers non-equipment / environmental factors that may affect product quality during the production process, forming a comprehensive traceability data source.

[0086] Next, anomalies are identified in the data from other monitoring areas of each production line. By comparing real-time data with pre-set process standards or historical data distribution models (such as acceptable ranges for raw material composition and normal fluctuation ranges for process parameters), anomalies are identified. For example, if the sulfur content of raw materials on a production line exceeds the standard upper limit of 0.05%, or if operators on a certain shift fail to follow standard procedures, these are flagged as other anomaly features, generating the first set of other anomaly features. Similarly, other anomaly feature sets for other production lines are generated.

[0087] Next, traceability analysis is performed on the set of other abnormal features and the set of product abnormal features. Data mining algorithms (such as the Apriori association rule, using the same process as above) are used to analyze the correlation between other abnormal features and product abnormal features. For example, through correlation analysis, the confidence level of the correlation between excessive sulfur content in batch A of raw materials and insufficient fracture strength of the product is determined with 90%. This generates a set of other causal factors for product abnormalities, clarifying the specific ways in which other factors influence product quality.

[0088] Finally, a graph is constructed to organize the abnormal feature sets and other traceable cause sets for each production line's products (similar to the previous steps, but will not be elaborated on due to the limitations of the manual). Using other abnormal features (such as process parameter adjustments and raw material batch anomalies) and product abnormal features as nodes, and causal relationships as directed edges, a third graph for tracing production line abnormalities is constructed. For example, the node for excessive sulfur content in raw materials in the graph points to the node for insufficient product fracture strength. The graph also annotates the association probability and historical frequency of occurrence, visually demonstrating the impact of non-equipment / environmental factors.

[0089] Through the process of multi-source data integration, intelligent detection of abnormal features, deep mining of causal relationships and graph visualization, it fills the gap in the existing technology for tracing other factors in the production process, significantly improves the efficiency and accuracy of locating complex anomalies in the manufacturing process, and provides comprehensive data support for the realization of multi-factor collaborative management and preventive process adjustments.

[0090] Furthermore, step A100 in the method provided in the embodiment of the present application includes: A110: Monitor each production line in real time to obtain a monitoring data set for each production line.

[0091] A120: Perform data cleaning based on the production line monitoring data sets to obtain monitoring standard data sets.

[0092] A130: Encapsulate and process the monitoring standard data sets according to the blockchain to obtain the production line monitoring blocks.

[0093] Alternatively, real-time monitoring can be implemented for each production line within an IoT production workshop. IoT sensors deployed throughout the production line (such as equipment status sensors, product feature collection devices, and environmental parameter monitoring devices) continuously collect multi-dimensional information, including equipment operating data, real-time product feature data, and ambient temperature and humidity. This generates raw monitoring datasets for each production line. These datasets contain a large amount of real-time dynamic data, such as the rotational speed of equipment on a particular production line, the frequency of temperature fluctuations, or real-time measurements of product dimensions, providing a fundamental data source for subsequent analysis.

[0094] Next, the original monitoring datasets for each production line are cleaned. Denoising algorithms are used to remove outliers (such as extreme data caused by temporary sensor failures), format standardization (unifying the data formats of different device interfaces), and missing value filling (using historical data or inferred data from adjacent nodes) to filter out invalid or erroneous data. This creates standardized monitoring datasets with unified structure and reliable quality. For example, abnormal points in the vibration data of equipment on a production line that exceed the normal fluctuation range are marked and corrected to ensure that the data truly reflects the actual production status.

[0095] Finally, blockchain technology is used to encapsulate the cleaned monitoring standard data sets. A hashing algorithm generates a unique hash value for each standard data set. This data is then packaged into blocks, combined with timestamp information, and linked to the blockchain in chronological order, creating tamper-proof and traceable monitoring blocks for each production line. The distributed ledger nature of blockchain ensures that once data is on-chain, it cannot be unilaterally modified. Furthermore, the traceability information for each block accurately locates the time, device, and original state of data collection.

[0096] Through real-time monitoring and collection of raw data, data cleaning to improve data quality, and blockchain encapsulation to ensure data credibility, full-chain management from production site data collection to trusted data storage is achieved, providing a reliable data foundation for subsequent product anomaly detection, multi-factor traceability, and anomaly management based on trusted data.

[0097] Furthermore, step A600 in the method provided in the embodiment of the present application includes: A610: Generate abnormality alarms for each production line according to the first production line abnormality tracing graph, the second production line abnormality tracing graph, and the third production line abnormality tracing graph.

[0098] In one embodiment, first, data from each production line monitoring block is collected in real time to obtain a set of abnormal characteristics of the current products of each production line (such as abnormal characteristics such as dimensional tolerance and surface cracks on the first production line).

[0099] Then, query the three traceability graphs separately to locate the root cause of the abnormal characteristics: In the equipment factor map, the associated edges of the dimensional deviation nodes were searched, and a causal relationship was found between the abnormal characteristics of the spindle speed fluctuation equipment and its confidence level of 95%, confirming the influence of the equipment factor; in the environmental factor map, it was verified whether there was a correlation between the abnormal environmental characteristics and the current product abnormality, such as the confidence level of 90% for temperature exceeding the limit and surface cracks; in other factor maps, it was checked whether there was a correlation between factors such as raw material batch fluctuations or process parameter adjustments and abnormal characteristics, such as the sulfur content of raw material A exceeding the standard → the product fracture strength was insufficient with a confidence level of 88%.

[0100] Combining the traceability results of the three graphs above, and using pre-set anomaly priority rules (such as equipment factors taking precedence over environmental factors), we can determine the primary anomaly root cause and impact path. For example, the primary cause of dimensional deviations on the first production line was spindle speed fluctuations (an equipment factor), with temperature exceeding the limit (an environmental factor) as a secondary cause.

[0101] Finally, a corresponding production line abnormality alert is generated based on the abnormality level (e.g., emergency, warning) and pushed to responsible departments such as equipment maintenance and process adjustment. This alert is accompanied by causal chain data from the traceability map (e.g., spindle speed fluctuation → dimensional deviation, 95% confidence level), enabling rapid response. The same steps are followed for other production lines, generating abnormality alerts for each line.

[0102] By integrating the causal correlation data of multi-dimensional traceability maps and combining them with real-time anomaly detection results, we have achieved full-process automated management from anomaly feature identification to root cause location, impact analysis and alarm generation, significantly improving the efficiency and accuracy of production line anomaly management and providing a scientific basis for real-time monitoring and preventive maintenance of the production process.

[0103] In summary, the IoT-driven real-time production process traceability monitoring method provided in the embodiments of the present application has the following technical effects: This application collects monitoring data sets of each production line in real time through monitoring equipment deployed on each production line in the IoT production workshop, and forms monitoring blocks for each production line through data cleaning and blockchain encapsulation. Through product feature recognition and production control solution reading, a product anomaly detection space is constructed and anomaly detection is performed to generate anomaly feature sets for each production line product. Combining the monitoring data of equipment, environment, and other factors, through causal analysis and graphical combing, a production line anomaly traceability map is constructed, and then anomaly alarms are generated to achieve accurate tracing and anomaly management of the production process, making manufacturing process management more intelligent and reliable, achieving the technical effect of accurately locating multiple factors such as equipment and environment that cause product anomalies and improving the comprehensiveness and timeliness of production anomaly management.

[0104] Example 2, as Figure 2 As shown, based on the same inventive concept as the aforementioned embodiment 1, the embodiment of the present application provides a real-time production process traceability monitoring system driven by the Internet of Things, the system comprising: The production line monitoring block acquisition module 1 is used to perform real-time monitoring on each production line in the IoT production workshop to obtain each production line monitoring block.

[0105] The abnormal feature set acquisition module 2 is used to perform product abnormality detection on each production line according to each production line monitoring block to obtain the abnormal feature set of each production line product.

[0106] The first map construction module 3 is used to trace the equipment factors of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and construct a first map for tracing the abnormality of the production line.

[0107] The second map construction module 4 is used to trace the environmental factors of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and construct a second map for tracing the abnormalities of the production line.

[0108] The third map construction module 5 is used to trace other factors of the abnormal feature set of each production line product according to each production line monitoring block, and construct a third map for tracing production line abnormalities.

[0109] The production line abnormality management module 6 is used to manage the production line abnormality according to the first production line abnormality tracing map, the second production line abnormality tracing map and the third production line abnormality tracing map.

[0110] Furthermore, the abnormal feature set acquisition module 2 is used to perform the following steps: Perform product feature identification according to the production line monitoring blocks to obtain product feature data of the first production line, product feature data of the second production line...product feature data of the P-th production line, where P is a positive integer; read the production control plan for each production line to obtain the production control plan for the first production line, the production control plan for the second production line...the production control plan for the P-th production line; perform product feature mining according to the production control plan for the first production line, the production control plan for the second production line...the production control plan for the P-th production line, and construct a first product anomaly detection space, a second product anomaly detection space...the P-th product anomaly detection space; perform anomaly detection on the product feature data of the first production line, the product feature data of the second production line...the product feature data of the P-th production line according to the first product anomaly detection space, the second product anomaly detection space...the P-th product anomaly detection space, respectively, to generate anomaly feature sets for the products of each production line.

[0111] Furthermore, the abnormal feature set acquisition module 2 is used to perform the following steps: Normal product features are retrieved according to the first production line production control plan to obtain a product feature record set; support evaluation is performed on the product feature record set to obtain a product feature support evaluation set; confidence analysis is performed on the product feature record set according to the product feature support evaluation set to obtain a product feature confidence evaluation set; based on the confidence evaluation constraints, the product feature record set is selected according to the product feature confidence evaluation set to obtain the first product anomaly detection space.

[0112] Furthermore, the first graph construction module 3 is used to perform the following steps: The abnormal feature sets of products on each production line include the first product abnormal feature set, the second product abnormal feature set...the Pth product abnormal feature set; equipment monitoring data is identified according to the monitoring blocks of each production line to obtain the first production line equipment monitoring area, the second production line equipment monitoring area...the Pth production line equipment monitoring area; abnormalities are identified according to the first production line equipment monitoring area, the second production line equipment monitoring area...the Pth production line equipment monitoring area to obtain the first equipment abnormal feature set, the second equipment abnormal feature set...the Pth equipment abnormal feature set; the first product abnormal feature set, the second product abnormal feature set...the Pth product abnormal feature set are respectively traced and analyzed according to the first equipment abnormal feature set, the second equipment abnormal feature set...the Pth equipment abnormal feature set to obtain multiple product abnormal equipment tracing sets; the abnormal feature sets of products on each production line and the multiple product abnormal equipment tracing sets are graphically combed to generate the first graph for tracing the production line abnormality.

[0113] Furthermore, the first graph construction module 3 is used to perform the following steps: Extract the Kth feature of the product abnormality according to the first product abnormality feature set, where K is a positive integer; extract associated data from the first device abnormality feature set according to the Kth feature of the product abnormality to obtain the Kth associated device abnormality feature; perform supervised training according to the product abnormality device tracing record set to obtain multiple product abnormality device tracing models; establish a first tracing node model and a second tracing node model based on the multiple product abnormality device tracing models; perform reinforcement learning fusion on the second tracing node model according to the first tracing node model to obtain a product abnormality device tracing channel; input the Kth feature of the product abnormality and the Kth associated device abnormality feature into the product abnormality device tracing channel to obtain a Kth product abnormality device tracing result, and add the Kth product abnormality device tracing result to the first product abnormality device tracing set.

[0114] Furthermore, the second map construction module 4 is used to perform the following steps: Environmental monitoring data is identified according to the production line monitoring blocks to obtain the first production line environmental monitoring area, the second production line environmental monitoring area...the Pth production line environmental monitoring area; anomalies are identified according to the first production line environmental monitoring area, the second production line environmental monitoring area...the Pth production line environmental monitoring area to obtain the first environmental anomaly feature set, the second environmental anomaly feature set...the Pth environmental anomaly feature set; based on the first environmental anomaly feature set, the second environmental anomaly feature set...the Pth environmental anomaly feature set, the first product anomaly feature set, the second product anomaly feature set...the Pth product anomaly feature set are respectively traced and analyzed to obtain multiple product anomaly environmental tracing sets; the product anomaly feature sets of each production line and the multiple product anomaly environmental tracing sets are graphically combed to generate the second graph for tracing the production line anomaly.

[0115] Furthermore, the third map construction module 5 is used to perform the following steps: Other monitoring data is identified based on the monitoring blocks of each production line to obtain the other monitoring areas of the first production line, the other monitoring areas of the second production line...the other monitoring areas of the P-th production line; abnormalities are identified based on the other monitoring areas of the first production line, the other monitoring areas of the second production line...the other monitoring areas of the P-th production line to obtain the first other abnormal feature set, the second other abnormal feature set...the P-th other abnormal feature set; the first product abnormal feature set, the second product abnormal feature set...the P-th product abnormal feature set are respectively traced and analyzed based on the first other abnormal feature set, the second other abnormal feature set...the P-th other abnormal feature set to obtain multiple product abnormality other tracing cause sets; the product abnormality feature sets of each production line and the multiple product abnormality other tracing cause sets are graphically combed to generate the third graph for tracing the production line abnormality.

[0116] Furthermore, the production line monitoring block acquisition module 1 is configured to perform the following steps: The production lines are monitored in real time to obtain monitoring data sets of each production line; data cleaning is performed based on the monitoring data sets of each production line to obtain monitoring standard data sets; and the monitoring standard data sets are packaged and processed according to the blockchain to obtain monitoring blocks of each production line.

[0117] Furthermore, the production line abnormality management module 6 is configured to perform the following steps: An abnormality alarm for each production line is generated according to the first production line abnormality tracing graph, the second production line abnormality tracing graph, and the third production line abnormality tracing graph.

[0118] The IoT-driven real-time production process traceability monitoring system provided in the embodiment of the present invention can execute the IoT-driven real-time production process traceability monitoring method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0119] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0120] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A real-time production process traceability monitoring method driven by the Internet of Things, characterized in that: The method comprises: Real-time monitoring of each production line in the IoT production workshop is carried out to obtain monitoring blocks for each production line; Perform product anomaly detection on each production line according to each production line monitoring block to obtain an abnormal feature set of products on each production line; Conducting equipment factor tracing of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and constructing a first graph for tracing the abnormality of the production line; Conduct environmental factor tracing of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and construct a second production line abnormality tracing graph; Tracing other factors of the abnormal feature set of the products of each production line according to the monitoring blocks of each production line to construct a third graph for tracing the abnormality of the production line; Production line abnormality management is performed according to the first production line abnormality tracing map, the second production line abnormality tracing map, and the third production line abnormality tracing map.

2. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: Perform product anomaly detection on each production line according to each production line monitoring block to obtain an abnormal feature set of products on each production line, including: Perform product feature recognition according to the monitoring blocks of each production line to obtain product feature data of the first production line, product feature data of the second production line, ... product feature data of the Pth production line, where P is a positive integer; Reading the production control plan for each of the production lines to obtain the production control plan for the first production line, the production control plan for the second production line, and so on, the production control plan for the Pth production line; Perform product feature mining based on the first production line production control plan, the second production line production control plan, ..., the Pth production line production control plan, to construct a first product anomaly detection space, a second product anomaly detection space, ..., a Pth product anomaly detection space; According to the first product anomaly detection space, the second product anomaly detection space...the Pth product anomaly detection space, anomaly detection is performed on the first production line product feature data, the second production line product feature data...the Pth production line product feature data respectively to generate the anomaly feature set of each production line product.

3. The real-time production process traceability monitoring method driven by the Internet of Things according to claim 2, characterized in that: Product feature mining is performed according to the first production line production control plan, the second production line production control plan, ..., the Pth production line production control plan, including: Perform normal product feature retrieval according to the first production line production control plan to obtain a product feature record set; Perform support evaluation based on the product feature record set to obtain a product feature support evaluation set; Performing confidence analysis on the product feature record set according to the product feature support evaluation set to obtain a product feature confidence evaluation set; Based on the confidence evaluation constraint, the product feature record set is selected according to the product feature confidence evaluation set to obtain the first product anomaly detection space.

4. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: The device factor tracing is performed on the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and a first graph for tracing the abnormality of the production line is constructed, including: The abnormal feature set of each production line product includes a first product abnormal feature set, a second product abnormal feature set ... a Pth product abnormal feature set; Identify equipment monitoring data according to each production line monitoring block to obtain a first production line equipment monitoring area, a second production line equipment monitoring area, ... a Pth production line equipment monitoring area; performing abnormality identification based on the first production line equipment monitoring area, the second production line equipment monitoring area, ... the Pth production line equipment monitoring area to obtain a first equipment abnormality feature set, a second equipment abnormality feature set, ... the Pth equipment abnormality feature set; performing tracing analysis on the first product abnormality feature set, the second product abnormality feature set, ... the Pth product abnormality feature set respectively based on the first device abnormality feature set, the second device abnormality feature set, ... the Pth device abnormality feature set, to obtain a plurality of product abnormality device tracing cause sets; The abnormal feature sets of the products of each production line and the multiple product abnormal equipment tracing cause sets are graphically sorted to generate the first graph of the production line abnormality tracing.

5. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 4, characterized in that: The first product abnormality feature set, the second product abnormality feature set, ..., the Pth product abnormality feature set are respectively subjected to causal analysis based on the first device abnormality feature set, the second device abnormality feature set, ..., the Pth device abnormality feature set, including: Extracting the Kth feature of product anomaly based on the first product anomaly feature set, where K is a positive integer; Extracting associated data from the first device abnormality feature set based on the Kth feature of the product abnormality to obtain the Kth associated device abnormality feature; Conduct supervised training based on the product abnormal equipment tracing record set to obtain multiple product abnormal equipment tracing models; Establishing a first tracing node model and a second tracing node model based on the multiple product abnormality equipment tracing models; Perform reinforcement learning fusion on the second tracing node model according to the first tracing node model to obtain a product abnormality equipment tracing channel; The Kth feature of the product abnormality and the Kth associated device abnormality feature are input into the product abnormal device tracing channel to obtain the Kth product abnormal device tracing result, and the Kth product abnormal device tracing result is added to the first product abnormal device tracing set.

6. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: Tracing the abnormal feature set of each production line product based on the environmental factors according to each production line monitoring block to construct a second production line abnormality tracing graph, including: Identify environmental monitoring data according to each production line monitoring block to obtain a first production line environmental monitoring area, a second production line environmental monitoring area, ... a Pth production line environmental monitoring area; performing abnormality identification based on the first production line environment monitoring area, the second production line environment monitoring area, ... the Pth production line environment monitoring area to obtain a first environmental abnormality feature set, a second environmental abnormality feature set, ... a Pth environmental abnormality feature set; performing causal analysis on the first product abnormality feature set, the second product abnormality feature set, ... the Pth product abnormality feature set respectively based on the first environmental abnormality feature set, the second environmental abnormality feature set, ... the Pth environmental abnormality feature set, to obtain multiple product abnormality environmental causal sets; The abnormal feature sets of the products of each production line and the abnormal environmental cause sets of the multiple products are graphically sorted to generate the second graph of the production line abnormality tracing.

7. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: According to the monitoring blocks of each production line, other factors of the abnormal feature set of the products of each production line are traced back to construct a third graph of production line abnormality tracing, including: Identify other monitoring data based on the monitoring blocks of each production line to obtain other monitoring areas of the first production line, other monitoring areas of the second production line, and so on and so forth; performing abnormality identification based on the first production line other monitoring area, the second production line other monitoring area, ... the Pth production line other monitoring area to obtain a first other abnormality feature set, a second other abnormality feature set, ... a Pth other abnormality feature set; performing causal analysis on the first product abnormality feature set, the second product abnormality feature set, ... the Pth product abnormality feature set respectively based on the first other abnormality feature set, the second other abnormality feature set, ... the Pth other abnormality feature set, to obtain multiple product abnormality other causal sets; The abnormal feature sets of the products of each production line and the other traceability sets of the multiple product abnormalities are graphically sorted to generate the third graph of the production line abnormality tracing.

8. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: Based on the real-time monitoring of each production line in the IoT production workshop, the monitoring blocks of each production line are obtained, including: Performing real-time monitoring on each of the production lines to obtain a monitoring data set for each production line; Performing data cleaning based on the production line monitoring data sets to obtain each monitoring standard data set; The monitoring standard data sets are encapsulated and processed according to the blockchain to obtain the production line monitoring blocks.

9. The method for real-time production process traceability monitoring driven by the Internet of Things according to claim 1, characterized in that: An abnormality alarm for each production line is generated according to the first production line abnormality tracing graph, the second production line abnormality tracing graph, and the third production line abnormality tracing graph.

10. The real-time production process traceability monitoring system driven by the Internet of Things is characterized by: A system for implementing the IoT-driven real-time production process traceability monitoring method according to any one of claims 1 to 9, comprising: The production line monitoring block acquisition module is used to monitor each production line in the IoT production workshop in real time and obtain the monitoring blocks of each production line; An abnormal feature set acquisition module, configured to perform product abnormality detection on each production line according to each production line monitoring block, and obtain an abnormal feature set of products on each production line; A first graph construction module is configured to perform equipment factor tracing on the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and construct a first graph for tracing the abnormality of the production line; A second graph construction module is configured to perform environmental factor tracing on the abnormal feature set of the products of each production line according to the monitoring blocks of each production line, and construct a second graph for tracing the abnormality of the production line; A third graph construction module is configured to perform other factor tracing on the abnormal feature set of each production line product according to each production line monitoring block, and construct a third production line abnormality tracing graph; The production line abnormality management module is used to manage the production line abnormality according to the first production line abnormality tracing map, the second production line abnormality tracing map and the third production line abnormality tracing map.

Images