Safety tool intelligent management full-life-cycle tracing system based on Internet of Things

Abstract

The invention discloses a safety tool intelligent management full life cycle tracing system based on the Internet of Things, and relates to the technical field of warehouse management, and the system comprises an Internet of Things data collection module which is used for obtaining environment parameters and use behavior data of tools; the behavior feature extraction module is used for extracting features to obtain a behavior feature sequence; the risk behavior mining module is used for counting risk behavior characteristics of the faulty tools and instruments based on the behavior characteristic sequence, defining a risk behavior graph, and learning embedded vectors and attention weights of risk behaviors by using a graph convolutional network; the risk level prediction module is used for encoding the real-time risk behavior characteristics by using the embedded vector and the attention weight, and predicting the risk level of the encoded real-time risk behavior characteristics by using an LSTM model; and the risk tracing and updating module is used for generating an event chain and an early warning log and updating a risk behavior graph and an LSTM model when the risk level is greater than or equal to a risk level threshold value, so that potential tool use risks can be found in advance.

Description

Technical Field

[0001] The present application relates to the field of warehouse management technology, and in particular to an intelligent management and full life cycle traceability system for safety tools based on the Internet of Things. Background Art

[0002] The full life cycle management of safety tools consists of a vertical interactive platform and a sub-cabinet, equipped with an intelligent management system for safety tools. It covers multiple key areas such as storage layout planning of safety tools in the cabinet, storage environment monitoring and control, and accurate recording of tool entry and exit information. It effectively solves the daily problems faced by safety tools, such as incomplete management, lack of standardized use, and inaccurate and unstable identification of tool information, and realizes the transformation from disorderly management to orderly and refined management.

[0003] A Chinese patent application with publication number CN116882896A discloses an identity-based inbound and outbound management system for power supply station tools. When the system is running, identity verification and identification ensure that only authorized personnel use the system, enhancing the security of the system. The distinction between different user roles and permission control effectively manage user permissions, ensuring the integrity and controllability of system operations. The accuracy and reliability of data are improved through automated tool information acquisition. The extraction and association of relevant features help to further understand the properties and status of tools, support subsequent analysis and management decisions, and automated inbound and outbound operations reduce the risk of human errors and inaccurate information, improve the reliability and accuracy of data, and then generate inbound and outbound documents and update tool information, which helps to track and record the status and location of tools in real time, providing traceability and management convenience.

[0004] However, in the existing safety tool management system, if a tool is damaged or an accident occurs, the system can only analyze the chain of responsibility after the fact and cannot detect potential risks in advance, resulting in the lack of a prevention mechanism. Summary of the Invention

[0005] This application aims to solve, at least to some extent, one of the technical problems in the related art. To this end, one purpose of this application is to propose an IoT-based intelligent management and full life cycle traceability system for safety tools, so as to identify potential tool usage risks in advance and effectively prevent safety accidents.

[0006] One aspect of the present application provides an IoT-based intelligent management and full lifecycle traceability system for safety tools, including:

[0007] IoT data collection module, used to obtain environmental parameters and tool usage behavior data;

[0008] Behavior feature extraction module, used to extract features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence;

[0009] The risk behavior mining module is used to statistically analyze the risk behavior characteristics of faulty tools based on the behavior feature sequence. It defines a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges. It uses a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph.

[0010] The risk level prediction module is used to encode real-time risk behavior features using embedding vectors and attention weights, and predict the risk level of the encoded real-time risk behavior features using an LSTM model;

[0011] The risk retrospective update module is used to set the risk level threshold. When the risk level is greater than or equal to the risk level threshold, an event chain and warning log are generated. The risk behavior graph and LSTM model are updated according to the warning log.

[0012] The specific method for obtaining the environmental parameters and tool usage behavior data is as follows:

[0013] Sensors are installed on each tool to collect environmental parameters and usage behavior data; the environmental parameters include temperature data, humidity data and location data; the usage behavior data includes usage frequency, usage duration and vibration data.

[0014] The specific method of extracting features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence is as follows:

[0015] Step S210: setting the time window length, counting the number of times used in each time window, and obtaining a usage frequency sequence;

[0016] Step S220: Count the N tools that are most frequently used by people in each time window to form a tool vector;

[0017] Step S230: Calculate the average time a person uses each tool in each time window based on the usage time to form a time matrix;

[0018] Step S240: Calculate the deviation between the actual trajectory of the tool and the standard trajectory in each time window;

[0019] Step S250: synthesizing the vibration data, calculating the vibration amplitude, setting a vibration threshold, and counting the proportion of vibration amplitudes greater than or equal to the vibration threshold in each time window;

[0020] Step S260: Setting a safe range for temperature and humidity, and counting the percentage of temperature and humidity exceeding the safe range in each time window;

[0021] Step S270: Using the extracted usage frequency sequence, tool vector, duration matrix, deviation, vibration amplitude ratio, temperature ratio, and humidity ratio as features extracted from the historical usage behavior data and environmental parameters to obtain a behavior feature segment for each time window;

[0022] Step S280: constructing a behavior feature sequence by combining the behavior feature segments of all time windows in chronological order.

[0023] The specific method of statistically analyzing the risk behavior characteristics of faulty tools based on the behavior characteristic sequence is as follows:

[0024] A historical accident dataset is constructed, wherein the historical accident dataset consists of a behavioral feature sequence of a faulty tool in an accident case within M time windows before the fault occurs and an accident label of the corresponding accident case; for each accident case, the frequent patterns of its behavioral feature sequence are extracted, and the frequent patterns of all accident cases are merged to obtain a frequent pattern set; in the frequent pattern set, a frequent pattern with a confidence level greater than or equal to a confidence threshold is selected as a risk behavior; and for each accident case, a risk behavior feature is generated, wherein the risk behavior feature consists of a behavioral feature sequence corresponding to the accident case and a sequence of indicator values ​​indicating whether the behavior feature sequence contains each risk behavior.

[0025] The specific method of defining a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges and using a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph is as follows:

[0026] Step S310: Define a risk behavior graph, with risk behaviors as nodes and risk behavior associations as edges. The attribute of the node is the probability of the risk behavior occurring in all accident cases, and the weight of the edge is the probability of two risk behaviors co-occurring in the same time window.

[0027] Step S320: Use a graph convolutional network to perform feature propagation and learning on the risk behavior graph to obtain an embedding vector for each risk behavior node;

[0028] Step S330: Use the attention mechanism to assign an attention weight to each risky behavior.

[0029] The specific method of encoding the real-time risk behavior feature using the embedding vector and the attention weight and predicting the risk level of the encoded real-time risk behavior feature using the LSTM model is as follows:

[0030] Step S410: Recording real-time usage behavior data and environmental parameters, and obtaining a real-time behavior feature sequence divided by time windows;

[0031] Step S420: determining whether the real-time behavior feature sequence contains each risk behavior to obtain a real-time indicator value sequence, and concatenating the real-time indicator value sequence and the real-time behavior feature sequence to obtain a real-time risk behavior feature;

[0032] Step S430: For each time window, encode the real-time indicator value sequence based on the embedding vector of the risk behavior and the attention weight to obtain the encoded real-time risk behavior feature;

[0033] Step S440: Use the encoded real-time risk behavior features to form a feature sequence, and predict the risk level of each time window based on the LSTM model.

[0034] The training process of the LSTM model is:

[0035] Step S441: Encoding the indicator values ​​in the risk behavior characteristics of the historical accident data set to obtain an encoded risk behavior characteristic sequence;

[0036] Step S442: Label the corresponding risk level labels for the coded risk behavior feature sequences of all accident cases, define a time window length f, use the first f-1 coded risk behavior features in the time window as input, and use the f-th risk level label in the time window as output. Combine the coded risk behavior feature sequence and the risk level label into training data;

[0037] The risk level labeling method is as follows: for each time window, a risk level label is generated according to whether an accident occurs in the subsequent time window. If an accident occurs in the subsequent q time windows, the risk level label is 1, otherwise the risk level label is 0;

[0038] Step S443: defining the structure of the LSTM model, including the input layer, the LSTM layer, and the output layer; using the binary cross entropy loss function as the loss function to measure the difference between the predicted risk level and the actual risk level label; minimizing the loss function using the backpropagation algorithm and the optimizer; and updating the LSTM model parameters;

[0039] Step S444: When the value of the loss function converges to the minimum, a trained LSTM model is obtained.

[0040] The risk level threshold is set, and when the risk level is greater than or equal to the risk level threshold, an event chain and a warning log are generated. The specific method for updating the risk behavior graph and the LSTM model according to the warning log is as follows:

[0041] Step S510: Setting a risk level threshold. When the risk level is greater than or equal to the risk level threshold, an early warning is triggered, and the system automatically sets the tool status to pending maintenance, prohibiting further lending.

[0042] Step S520: extracting the indicator value sequence of the risk behavior of the tool in the latest f time windows, associating the risk behavior with personnel, environment, and tasks, and generating an event chain;

[0043] Step S530: Record the time of each warning, the warning tools, personnel, the indicator value sequence of the risk behavior, and the investigation results, generate a warning log, regularly update the risk behavior graph based on the warning log, and recalculate the embedding vectors and attention weights of the nodes in the risk behavior graph based on the updated risk behavior graph;

[0044] Step S540: Using the updated embedding vector and attention weight, the real-time risk behavior feature is encoded and updated;

[0045] Step S550: Add the encoded risk behavior feature sequences of the accident cases with confirmed risks in the early warning log to the training data, and regularly update the LSTM model parameters.

[0046] One aspect of the present application provides a full lifecycle traceability method for intelligent management of safety tools based on the Internet of Things, including:

[0047] Step S100: Acquire environmental parameters and tool usage behavior data;

[0048] Step S200: extracting features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence;

[0049] Step S300: Counting risk behavior features of faulty tools based on the behavior feature sequence, defining a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges, and using a graph convolutional network to learn embedding vectors and attention weights of risk behaviors in the risk behavior graph;

[0050] Step S400: Encode the real-time risk behavior feature using the embedding vector and attention weight, and predict the risk level of the encoded real-time risk behavior feature using the LSTM model;

[0051] Step S500: Set a risk level threshold, generate an event chain and a warning log when the risk level is greater than or equal to the risk level threshold, and update the risk behavior graph and LSTM model according to the warning log.

[0052] The IoT-based intelligent management and full life cycle traceability system for safety tools proposed in this application has the following advantages over existing technologies:

[0053] This application deploys various sensors on tools and leverages IoT technology to automatically collect, remotely transmit, and cloud-based store tool data. By setting a time window, continuous tool data is discretized and behavioral features are extracted from multiple dimensions, resulting in a structured behavioral feature sequence that provides a suitable data representation for risk behavior analysis.

[0054] This application uses frequent pattern mining technology to mine risk behaviors that lead to accidents from massive historical accident data, revealing the behavioral root causes of accidents and providing key focus for risk prevention and control.

[0055] This application constructs a risk behavior graph, uses the mined risk behaviors as nodes, and the associations between behaviors as edges, to construct a global risk behavior map and systematically analyze the association network of risk behaviors.

[0056] This application uses graph convolutional network technology to reveal the inherent mechanism of risk evolution and automatically learn the deep characteristics of risk behavior; at the same time, through the attention mechanism, it learns the importance differences of different risk behaviors and provides a new method for quantitative risk assessment.

[0057] This application applies the embedding vectors and attention weights of risk behaviors to the encoding of real-time data, highlighting the impact of key risk behaviors, and obtaining a risk feature representation that comprehensively considers the correlation and importance of risk behaviors, providing information-rich feature input for risk prediction. It also uses the LSTM model to predict risk levels, fully exploring the time-dependent laws of risk behaviors, predicting risk evolution trends, and realizing dynamic risk estimation and trend warning.

[0058] This application provides data support for accident cause analysis and responsibility tracing by generating event chains and early warning logs, moving safety management forward, achieving early detection, early prevention, and early disposal of risks, and improving the timeliness and refinement of safety management.

[0059] This application introduces an online update mechanism for models and knowledge, giving the system the ability to self-learn and continuously evolve, realizing a dynamic closed loop of "knowledge-data-knowledge", adapting to changes in the tool usage environment, and continuously maintaining the effectiveness and accuracy of risk warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A timing diagram of the IoT-based intelligent management and full life cycle traceability system for safety tools provided in this application;

[0061] Figure 2 Flowchart for mining risk behavior features provided for this application;

[0062] Figure 3A schematic diagram of the structure of the risk behavior graph provided for this application;

[0063] Figure 4 This is a flow chart of the method for tracing the entire life cycle of intelligent management of safety tools based on the Internet of Things provided in this application. DETAILED DESCRIPTION

[0064] To better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely descriptions of exemplary embodiments of the present application and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0065] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The accompanying drawings are for illustration only and are not drawn strictly to scale. As used herein, the terms "substantially," "approximately," and similar terms are used to indicate approximate values, not degrees, and are intended to illustrate inherent deviations in measurements or calculations that would be recognized by a person of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these steps would occur in actual operation, unless otherwise specified or inferred from the context.

[0066] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0067] Unless otherwise defined, all terms used herein (including engineering and scientific terms) have the same meaning as commonly understood by those skilled in the art to which this application belongs. It should also be understood that, unless otherwise specified in this application, words defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0068] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0069] Example 1:

[0070] like Figure 1 As shown, the IoT-based intelligent management and full life cycle traceability system for safety tools provided in this application includes:

[0071] IoT data collection module, used to obtain environmental parameters and tool usage behavior data;

[0072] The specific method for obtaining the environmental parameters and tool usage behavior data is as follows:

[0073] Install sensors on each tool to collect environmental parameters and usage behavior data; the environmental parameters include temperature data, humidity data, and location data; the usage behavior data includes usage frequency, usage duration, and vibration data;

[0074] Usage frequency means recording the number of times a user uses the device within a certain period of time through a counter.

[0075] Usage time refers to the time a tool is used, which is obtained by recording the start and end time of each use through a timer.

[0076] The vibration data refers to the three-axis vibration data collected by the acceleration sensor.

[0077] The temperature data represents the ambient temperature collected by the temperature sensor.

[0078] Humidity data represents the ambient humidity collected by the humidity sensor.

[0079] The location data represents the location coordinates collected by GPS or an indoor positioning module.

[0080] Furthermore, the environmental parameters and tool usage behavior data are sampled at a fixed frequency by setting a sampling interval;

[0081] Furthermore, the collected environmental parameters and usage behavior data are uploaded to the cloud server in real time via RFID to achieve remote transmission and centralized storage of data.

[0082] This application installs sensors on each tool to collect real-time usage data and environmental parameters. The sensors sample at a fixed frequency to ensure the continuity of the collected data. IoT-based data collection can greatly improve the timeliness and reliability of data, providing a data foundation for subsequent feature extraction, data analysis, and risk assessment.

[0083] Behavior feature extraction module, used to extract features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence;

[0084] The specific method of extracting features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence is as follows:

[0085] Step S210: setting the time window length, counting the number of times used in each time window, and obtaining a usage frequency sequence;

[0086] The time window length is , the number of sampling points contained in each time window is , the usage frequency in M ​​time windows is counted, and the obtained usage frequency sequence is expressed as ,in, is the usage frequency in the mth time window, m=1,2,...,M, is the single sampling duration, Indicates rounding down;

[0087] Step S220: Count the N tools that are most frequently used by people in each time window to form a tool vector;

[0088] The tool vector is represented as ,in, represents the nth tool most frequently used by the personnel, N represents the number of tools most frequently used, n=1,2,...,N;

[0089] Step S230: Calculate the average time a person uses each tool in each time window based on the usage time to form a time matrix;

[0090] The duration matrix ,in, represents the time matrix of personnel using each tool in the mth time window, Indicates personnel Use tools The average duration of

[0091] Step S240: Calculate the deviation between the actual trajectory of the tool and the standard trajectory in each time window;

[0092] The degree of deviation The calculation uses the dynamic time warping algorithm;

[0093] Step S250: synthesizing the vibration data, calculating the vibration amplitude, setting a vibration threshold, and counting the proportion of vibration amplitudes greater than or equal to the vibration threshold in each time window;

[0094] The vibration data middle 、 、 Respectively represents the vibration data on the x, y, and z axes; vibration threshold It is set by those skilled in the art based on experience.

[0095] The vibration amplitude The calculation formula is: ;

[0096] The calculation formula of the vibration amplitude ratio is: ,in, represents the proportion of vibration amplitude in the mth time window, is the number of sampling points greater than or equal to the vibration threshold;

[0097] Step S260: Setting a safe range for temperature and humidity, and counting the percentage of temperature and humidity exceeding the safe range in each time window;

[0098] The calculation formula of the temperature ratio is: ,in, is the percentage of temperatures exceeding the safe temperature range in the mth time window, The number of sampling points that exceed the safe temperature range;

[0099] The calculation formula of the humidity ratio is: ,in, is the percentage of humidity that exceeds the safe range of humidity in the mth time window, The number of sampling points that exceed the safe range of humidity;

[0100] Step S270: Using the extracted usage frequency sequence, tool vector, duration matrix, deviation, vibration amplitude ratio, temperature ratio, and humidity ratio as features extracted from the historical usage behavior data and environmental parameters to obtain a behavior feature segment for each time window;

[0101] The behavior feature segment can be expressed as ;

[0102] Step S280: constructing a behavior feature sequence by combining the behavior feature segments of all time windows in chronological order.

[0103] The above steps extract features from the collected historical data. By setting a time window, the continuous time series data is divided into discrete time segments, each of which corresponds to a behavioral feature segment. This time window-based feature extraction method in this application not only considers the temporal dependence of the data but also facilitates subsequent feature analysis. The extracted behavioral features comprehensively depict the usage status and environment of the tool, providing data support for subsequent risk behavior mining.

[0104] The risk behavior mining module is used to statistically analyze the risk behavior characteristics of faulty tools based on the behavior feature sequence. It defines a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges. It uses a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph.

[0105] like Figure 2 As shown, this is a flowchart for mining risk behavior features provided by the present application. The specific method for statistically analyzing risk behavior features of faulty tools based on behavior feature sequences is as follows: constructing a historical accident data set, wherein the historical accident data set consists of a behavior feature sequence of the faulty tools in the accident case within M time windows before the fault occurs and an accident label of the corresponding accident case; for each accident case, extracting frequent patterns of its behavior feature sequence, merging the frequent patterns of all accident cases to obtain a frequent pattern set; selecting frequent patterns with a confidence level greater than or equal to a confidence threshold in the frequent pattern set as risk behaviors; generating a risk behavior feature for each accident case, wherein the risk behavior feature consists of a behavior feature sequence corresponding to the accident case and a sequence of indicator values ​​indicating whether each risk behavior is contained in the behavior feature sequence;

[0106] The historical accident dataset can be expressed as , where G is the total number of accident cases, is the behavioral feature sequence of the faulty tool in the g-th accident case within the M time windows before the accident, where represents the behavioral feature segment in the mth time window before the gth accident case, m=1,2,...,M, For the accident label, =0 means that no accident occurred in the g-th accident case, =1 means that an accident occurred in the g-th accident case.

[0107] The frequent pattern refers to a combination of behavioral features whose support is greater than or equal to a support threshold in the behavioral feature sequence of an accident case; the frequent pattern is extracted using a data mining algorithm; preferably, the data mining algorithm adopts an Apriori algorithm;

[0108] The calculation formula of the support is: ,in, Indicates the combination of behavioral features for which support needs to be calculated. Indicates the uth behavioral feature segment in the behavioral feature combination whose support needs to be calculated, Indicates the number of times the behavior feature combination appears in the behavior feature sequence; the support threshold is 0.1~0.5;

[0109] The frequent pattern can be expressed as ,in, represents the number of behavioral feature combinations with support greater than or equal to the support threshold in the behavioral feature sequence of the g-th accident case, Indicates the first frequent pattern A combination of behavioral characteristics;

[0110] The frequent pattern set can be expressed as: ;

[0111] The confidence level The calculation formula is: ;in To include frequent patterns Number of accident cases;

[0112] The confidence threshold is 0.5-0.9, and the frequent patterns with confidence greater than or equal to the confidence threshold are selected as risk behaviors, which are recorded as , K represents the total number of risk behaviors mined from the historical accident data set, represents the K-th risky behavior;

[0113] The risk behavior characteristics are expressed as ,in, is the indicator value sequence of whether the behavior feature sequence contains each risk behavior, , Indicates whether the behavior feature sequence contains the k-th risk behavior, k=1,2,...,K, , =0 means that the behavior feature sequence does not contain the k-th risk behavior, =1 indicates that the behavior feature sequence contains the kth risk behavior.

[0114] The above steps provide data support and explainability for the embedding vector learning of graph convolutional networks by mining the correlation patterns of risky behaviors from historical accident datasets for subsequent risk assessment.

[0115] The specific method of defining a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges and using a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph is as follows:

[0116] Step S310: Define a risk behavior graph, with risk behaviors as nodes and risk behavior associations as edges. The attribute of the node is the probability of the risk behavior occurring in all accident cases, and the weight of the edge is the probability of two risk behaviors co-occurring in the same time window.

[0117] like Figure 3 , which is a schematic diagram of the structure of the risk behavior graph provided by the present application, wherein the risk behavior graph is a graph structure consisting of nodes and edges between nodes in the graph;

[0118] The calculation formula of the node attribute is: ,in, is the attribute of the node corresponding to the ki-th risk behavior, The indicator value indicating whether the m-th time window of the g-th accident case contains the ki-th risk behavior;

[0119] The calculation formula of the edge weight is: ,in, is the weight of the edge between the nodes corresponding to the ki-th and kj-th risk behaviors, The indicator value indicating whether the m-th time window of the g-th accident case contains the kj-th risk behavior;

[0120] Step S320: Use a graph convolutional network to perform feature propagation and learning on the risk behavior graph to obtain an embedding vector for each risk behavior node;

[0121] The graph convolutional network includes an input layer, a convolutional layer, and an output layer. The adjacency matrix and node feature matrix of the risk behavior graph are input into the input layer. The convolutional layer is composed of multiple graph convolutional layers stacked together. Each graph convolutional layer forward propagates the risk behavior features of the node to obtain the embedded representation of the node. The output layer is the last graph convolutional layer, which outputs the final embedding vector of the node.

[0122] Step S330: using the attention mechanism to assign an attention weight to each risky behavior;

[0123] The calculation formula of the attention weight is: ,in, 、 Represents risky behavior and risky behaviors The attention score, 、 represents the ki and kj risk behaviors, is an exponential function;

[0124] The calculation formula of the attention score is: v. 、 are the learnable parameters of the graph convolutional network, representing the attention vector, weight matrix and bias term respectively, is the transpose of the attention vector, Risky behavior The embedding vector of is the activation function.

[0125] In the above steps, the graph convolutional network propagates and aggregates features of the risk behavior graph, learns the associations and implicit patterns between risk behaviors, finds out risk behaviors that may lead to accidents, provides a more comprehensive and in-depth feature representation for risk assessment, effectively mines the risk behavior patterns in historical accident data sets, and improves the accuracy and sensitivity of risk perception. Then, a risk behavior graph is constructed with risk behaviors as nodes, and the graph convolutional network is used to perform feature learning on the risk behavior graph to obtain an embedding vector for each risk behavior node. The embedding vector contains the semantic information and contextual relationship of the risk behavior. In addition, this application also uses the attention mechanism to assign attention weights to each risk behavior, highlighting the impact of key risk behaviors, making full use of the representation capabilities of the graph convolutional network, and automatically learning the risk behavior patterns in the tool use process, greatly improving the intelligence level of risk assessment.

[0126] The risk level prediction module is used to encode real-time risk behavior features using embedding vectors and attention weights, and predict the risk level of the encoded real-time risk behavior features using an LSTM model;

[0127] The specific method of encoding the real-time risk behavior feature using the embedding vector and the attention weight and predicting the risk level of the encoded real-time risk behavior feature using the LSTM model is as follows:

[0128] Step S410: Recording real-time usage behavior data and environmental parameters, and obtaining a real-time behavior feature sequence divided by time windows;

[0129] The real-time behavior feature sequence can be expressed as ,in, represents the behavior feature segment in the mth time window in the real-time behavior feature sequence, m=1,2,...,M;

[0130] Step S420: determining whether the real-time behavior feature sequence contains each risk behavior to obtain a real-time indicator value sequence, and concatenating the real-time indicator value sequence and the real-time behavior feature sequence to obtain a real-time risk behavior feature;

[0131] The real-time risk behavior characteristics can be expressed as , where b is the real-time indicator value sequence of whether the real-time behavior feature sequence contains each risk behavior, , Indicates whether the real-time behavior feature sequence contains the indicator value of the ki-th risk behavior, ki=1,2,...,K, , =0 means that the real-time behavior feature sequence does not contain the ki-th risk behavior, =1 means that the real-time behavior feature sequence contains the ki-th risk behavior;

[0132] Step S430: For each time window, encode the real-time indicator value sequence based on the embedding vector of the risk behavior and the attention weight to obtain the encoded real-time risk behavior feature;

[0133] The encoded real-time risk behavior characteristics are expressed as ;in, is the indicator value of the ki-th risk behavior in the m-th time window;

[0134] Step S440: Using the encoded real-time risk behavior features to form a feature sequence, and predicting the risk level of each time window based on the LSTM model;

[0135] The risk level is expressed as , represents the risk level of the mth time window, when =1, indicating high risk, when =0, indicating no risk;

[0136] The training process of the LSTM model is:

[0137] Step S441: Encoding the indicator values ​​in the risk behavior characteristics of the historical accident data set to obtain an encoded risk behavior characteristic sequence;

[0138] The coded risk behavior feature sequence can be expressed as ;

[0139] Step S442: Label the corresponding risk level labels for the coded risk behavior feature sequences of all accident cases, define a time window length f, use the first f-1 coded risk behavior features in the time window as input, and use the f-th risk level label in the time window as output. Combine the coded risk behavior feature sequence and the risk level label into training data;

[0140] Preferably, the time window length f is 3;

[0141] The risk level labeling method is as follows: for each time window, a risk level label is generated according to whether an accident occurs in the subsequent time window. If an accident occurs in the subsequent q time windows, the risk level label is 1, otherwise the risk level label is 0;

[0142] Preferably, q=3.

[0143] Step S443: defining the structure of the LSTM model, including the input layer, the LSTM layer, and the output layer; using the binary cross entropy loss function as the loss function to measure the difference between the predicted risk level and the actual risk level label; minimizing the loss function using the backpropagation algorithm and the optimizer; and updating the LSTM model parameters;

[0144] Step S444: When the value of the loss function converges to the minimum, a trained LSTM model is obtained.

[0145] The above steps train the LSTM model on the encoded risk behavior feature sequence to learn the long-term dependencies and risk evolution laws in the time feature sequence, thereby predicting the risk level of each time window. The trained LSTM model is used to conduct real-time risk assessment on the real-time collected usage behavior data and environmental parameters to provide support for early warning and decision-making.

[0146] For example, when new usage behavior data and environmental parameters are uploaded to the cloud, their behavioral feature sequences are first extracted. Then, the risk behavior indicator values ​​are encoded using the risk behavior embedding vector and attention weights, resulting in a comprehensive real-time risk behavior feature. Finally, this encoded real-time risk behavior feature sequence is input into a pre-trained LSTM model. Leveraging the LSTM model's ability to automatically learn long-term dependencies between feature sequences, the risk level for each time window is predicted, capturing the temporal evolution of risk behavior and enabling automated and real-time risk assessment.

[0147] The risk retrospective update module is used to set the risk level threshold. When the risk level is greater than or equal to the risk level threshold, an event chain and warning log are generated. The risk behavior graph and LSTM model are updated based on the warning log.

[0148] The risk level threshold is set, and when the risk level is greater than or equal to the risk level threshold, an event chain and a warning log are generated. The specific method for updating the risk behavior graph and the LSTM model according to the warning log is as follows:

[0149] Step S510: Setting risk level threshold When the risk level is greater than or equal to the risk level threshold, an early warning is triggered and the system automatically sets the tool status to pending maintenance, prohibiting further lending;

[0150] The risk level threshold The value of is 1;

[0151] Step S520: extracting the indicator value sequence of the risk behavior of the tool in the latest f time windows, associating the risk behavior with personnel, environment, and tasks, and generating an event chain;

[0152] Step S530: Record the time of each warning, the warning tools, personnel, the indicator value sequence of the risk behavior, and the subsequent investigation results, generate a warning log, regularly update the risk behavior graph based on the warning log, and recalculate the embedding vectors and attention weights of the nodes in the risk behavior graph based on the updated risk behavior graph;

[0153] The screening results are divided into confirmed risks and false positives;

[0154] Step S540: Using the updated embedding vector and attention weight, the real-time risk behavior feature is encoded and updated;

[0155] Step S550: Add the encoded risk behavior feature sequences of the accident cases with confirmed risks in the early warning log to the training data, and regularly update the LSTM model parameters.

[0156] The above steps achieve continuous updating and optimization of the risk behavior graph and LSTM model through human-computer interaction and model update.

[0157] The IoT-based intelligent management and full lifecycle traceability system for safety tools in this application fully utilizes IoT technology for data collection, transmission, and storage, forming a complete data closed loop. Furthermore, data mining using IoT-based data, combined with graph neural networks and LSTM models, enables intelligent analysis and risk assessment of massive amounts of historical and real-time data, realizing an automated and optimizable intelligent management and traceability system for safety tools.

[0158] Example 2:

[0159] like Figure 4 As shown, the IoT-based intelligent management and full life cycle traceability method for safety tools provided in this application includes:

[0160] Step S100: Acquire environmental parameters and tool usage behavior data;

[0161] Step S200: extracting features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence;

[0162] Step S300: Counting risk behavior features of faulty tools based on the behavior feature sequence, defining a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges, and using a graph convolutional network to learn embedding vectors and attention weights of risk behaviors in the risk behavior graph;

[0163] Step S400: Encode the real-time risk behavior feature using the embedding vector and attention weight, and predict the risk level of the encoded real-time risk behavior feature using the LSTM model;

[0164] Step S500: Set a risk level threshold, generate an event chain and a warning log when the risk level is greater than or equal to the risk level threshold, and update the risk behavior graph and LSTM model according to the warning log.

[0165] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0166] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. The IoT-based intelligent management and full life cycle traceability system for safety tools is characterized by: include: IoT data collection module, used to obtain environmental parameters and tool usage behavior data; Behavior feature extraction module, used to extract features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence; The risk behavior mining module is used to statistically analyze the risk behavior characteristics of faulty tools based on the behavior feature sequence. It defines a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges. It uses a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph. The risk level prediction module is used to encode real-time risk behavior features using embedding vectors and attention weights, and predict the risk level of the encoded real-time risk behavior features using an LSTM model; The risk retrospective update module is used to set the risk level threshold. When the risk level is greater than or equal to the risk level threshold, an event chain and warning log are generated. The risk behavior graph and LSTM model are updated according to the warning log.

2. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 1 is characterized in that: The specific method for obtaining environmental parameters and tool usage behavior data is: installing sensors on each tool to collect environmental parameters and usage behavior data; the environmental parameters include temperature data, humidity data and location data; the usage behavior data includes usage frequency, usage duration and vibration data.

3. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 2 is characterized in that: The specific method of extracting features from historical usage behavior data and environmental parameters to obtain a behavior feature sequence is as follows: Set the time window length, count the number of times it is used in each time window, and obtain the frequency sequence of use; count the N tools that are most frequently used by personnel in each time window to form a tool vector; calculate the average time that personnel use each tool in each time window based on the usage time to form a time matrix; calculate the deviation between the actual trajectory of the tool and the standard trajectory in each time window; synthesize the vibration data, calculate the vibration amplitude, set the vibration threshold, and count the proportion of vibration amplitudes greater than or equal to the vibration threshold in each time window; set the safety range of temperature and humidity, and count the proportion of temperature and humidity that exceed the safety range in each time window; use the extracted frequency sequence, tool vector, time matrix, deviation, vibration amplitude ratio, temperature ratio, and humidity ratio as features extracted from historical usage behavior data and environmental parameters to obtain the behavioral feature segment of each time window; construct the behavioral feature sequence of all time windows in chronological order.

4. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 3 is characterized in that: The specific method for statistically analyzing the risk behavior characteristics of faulty tools based on behavior feature sequences is as follows: constructing a historical accident data set, wherein the historical accident data set consists of the behavior feature sequences of the faulty tools in the accident cases within M time windows before the occurrence of the fault and the accident labels of the corresponding accident cases; for each accident case, extracting the frequent patterns of its behavior feature sequence, merging the frequent patterns of all accident cases, and obtaining a frequent pattern set; selecting the frequent patterns with a confidence level greater than or equal to a confidence threshold in the frequent pattern set as risk behaviors; and generating the risk behavior characteristics for each accident case, wherein the risk behavior characteristics consist of the behavior feature sequence of the corresponding accident case and a sequence of indicator values ​​indicating whether the behavior feature sequence contains each risk behavior.

5. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 4 is characterized in that: The specific method of defining a risk behavior graph with risk behaviors as nodes and risk behavior associations as edges and using a graph convolutional network to learn the embedding vectors and attention weights of risk behaviors in the risk behavior graph is as follows: Define a risk behavior graph, with risk behaviors as nodes and risk behavior associations as edges. The attribute of the node is the probability of the risk behavior occurring in all accident cases, and the weight of the edge is the probability of two risk behaviors co-occurring in the same time window. Use graph convolutional networks to propagate and learn features of the risk behavior graph and obtain the embedding vector of each risk behavior node; Use the attention mechanism to assign attention weights to each risky behavior.

6. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 5 is characterized in that: The specific method of encoding the real-time risk behavior feature using the embedding vector and the attention weight and predicting the risk level of the encoded real-time risk behavior feature using the LSTM model is as follows: Record real-time usage behavior data and environmental parameters, and obtain real-time behavior feature sequences divided by time windows; Based on determining whether the real-time behavior feature sequence contains each risk behavior, a real-time indicator value sequence is obtained, and the real-time risk behavior feature is obtained by splicing the real-time indicator value sequence and the real-time behavior feature sequence; For each time window, the real-time indicator value sequence is encoded based on the embedding vector of the risk behavior and the attention weight to obtain the encoded real-time risk behavior feature; The encoded real-time risk behavior features are used to form a feature sequence, and the risk level of each time window is predicted based on the LSTM model.

7. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 6, characterized in that: The training process of the LSTM model is: Encoding the indicator values ​​in the risk behavior characteristics of the historical accident data set to obtain an encoded risk behavior characteristic sequence; The coded risk behavior feature sequences of all accident cases are labeled with corresponding risk level labels. The time window length f is defined. The first f-1 coded risk behavior features in the time window are used as input, and the f-th risk level label in the time window is used as output. The coded risk behavior feature sequence and risk level label are combined into training data. Define the structure of the LSTM model, including the input layer, LSTM layer, and output layer. Use the binary cross-entropy loss function as the loss function to measure the difference between the predicted risk level and the actual risk level label. Use the backpropagation algorithm and optimizer to minimize the loss function and update the LSTM model parameters. When the value of the loss function converges to the minimum, the trained LSTM model is obtained.

8. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 7, characterized in that: The risk level labeling method is as follows: for each time window, a risk level label is generated based on whether an accident occurs in its subsequent time window. If an accident occurs in the subsequent q time windows, the risk level label is 1, otherwise the risk level label is 0.

9. The IoT-based safety tool intelligent management and full life cycle traceability system according to claim 8, characterized in that: The risk level threshold is set, and when the risk level is greater than or equal to the risk level threshold, an event chain and an early warning log are generated. The specific method for updating the risk behavior graph and the LSTM model according to the early warning log is as follows: Set a risk level threshold. When the risk level is greater than or equal to the risk level threshold, an early warning is triggered and the system automatically sets the tool status to pending maintenance, prohibiting further lending. Extract the indicator value sequence of the risk behavior of the tool in the last f time windows, associate the risk behavior with personnel, environment, and tasks, and generate an event chain; Record the time of each warning, the warning tools, personnel, the indicator value sequence of the risk behavior, and the investigation results, generate a warning log, regularly update the risk behavior graph based on the warning log, and recalculate the embedding vectors and attention weights of the nodes in the risk behavior graph based on the updated risk behavior graph; Use the updated embedding vectors and attention weights to encode and update the real-time risk behavior features; The encoded risk behavior feature sequences of accident cases with confirmed risks in the early warning log are added to the training data, and the LSTM model parameters are updated regularly.

Images