Radar security equipment full life cycle health management system and method

Abstract

The invention provides a full-life-cycle health management system and method for radar security and protection equipment, belongs to the technical field of equipment monitoring, and solves the technical problems of low equipment health state precision and lack of security and protection maintenance decisions caused by incapability of monitoring the full life cycle of radar equipment in the prior art. The system comprises a data acquisition module used for acquiring and recording full life cycle data of equipment; the digital thread data processing module is used for constructing a time sequence feature sequence and carrying out data association to obtain a digital thread data chain; a health prediction module which carries a CNN time sequence prediction model and carries out health state prediction on the time sequence feature sequence; the full life cycle state management module is used for adjusting life cycle stages of the equipment; and the visual interaction and decision support module is used for displaying the real-time state of the equipment, performing visual analysis and generating a maintenance decision. The method is applied to health monitoring and maintenance decision making of radar security and protection equipment.

Description

Technical Field

[0001] This invention belongs to the field of equipment monitoring technology, and particularly relates to the field of health monitoring technology for radar security equipment, specifically applied to the health monitoring and maintenance decision-making of radar security equipment. Background Technology

[0002] With the large-scale application of radar security equipment in all-domain security scenarios, the equipment faces complex environmental factors such as high and low temperatures, high humidity, and vibration during long-term operation, leading to increased failure risks. Traditional radar health management solutions have several core shortcomings. First, they lack full lifecycle coverage. Existing solutions mostly focus on single-point fault monitoring during the operational phase, failing to cover the entire lifecycle from manufacturing to installation, operation, maintenance, replacement, and decommissioning, thus failing to achieve full traceability and dynamic management of equipment status. Second, they suffer from insufficient accuracy in predictive models. Current solutions rely on single-parameter threshold judgments or simple statistical models, failing to fully exploit the temporal correlation characteristics of multiple parameters such as signal strength, temperature, and vibration, resulting in low accuracy in health status prediction and delayed fault warnings. Third, maintenance decisions lack dynamism. Current maintenance plans are mostly formulated on a fixed schedule, without considering the real-time health status of the equipment, historical fault records, and lifecycle stages, leading to over-maintenance or under-maintenance. Summary of the Invention

[0003] In view of this, the present invention aims to propose a full life cycle health management system and method for radar security equipment, so as to solve the technical problems of low accuracy of equipment health status and lack of security maintenance decision-making in the prior art due to the inability to monitor the full life cycle of radar equipment.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: This invention proposes a full lifecycle health management system for radar security equipment, the system comprising: Data acquisition module: used to collect and record the entire lifecycle data of the device; the entire lifecycle data includes data of the device operation phase, as well as process record data throughout each stage of the entire lifecycle; The digital thread data processing module is used to preprocess the full lifecycle data to construct a time-series feature sequence; it is also used to perform data association based on timestamps to obtain a digital thread data chain; the digital thread data chain is traceable and used to backtrack and analyze historical data. Health prediction module: Equipped with a CNN time-series prediction model, the health status of the time-series feature sequence is predicted through a three-layer one-dimensional convolutional neural network of the model, the probability distribution of the health status is obtained, and the health status with the highest probability distribution is taken as the prediction result; the health status includes normal, attention, warning, alarm and fault. The full lifecycle status management module adjusts the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning. Visualization and Decision Support Module: Based on the digital thread data chain, it displays the real-time status of the device, performs visual analysis of the real-time status, and generates maintenance decisions.

[0005] Furthermore, in the data acquisition module, the operational phase data includes the radar equipment's intensity, temperature, humidity, vibration value, and operating current; the process recording data includes equipment manufacturing information, installation date, maintenance records, fault records, and life cycle stage change records.

[0006] Furthermore, in the digital thread data processing module, the method for constructing the time-series feature sequence is as follows: Normalization is performed using standardization methods: , Where μ is the mean of the parameters, The standard deviation of the parameter; Extract N consecutive collected samples and construct the time-series feature sequence as follows: , in, is the multi-parameter feature vector at the current time; N=20, which is the sequence length.

[0007] Furthermore, the CNN time-series prediction model is as follows: The model input is the time-series feature sequence, wherein the feature parameters include signal strength, temperature, humidity, vibration, and current; Three one-dimensional convolutional layers: the number of channels is 5, 32, 64 and 128 respectively, the kernel size of each layer is 3, and the padding is 1; Each one-dimensional convolutional layer is followed by a max pooling layer, which has a pooling kernel size of 2 and a stride of 2, and is used for feature dimensionality reduction. The activation function used is the ReLU function; Dropout with a probability of 0.5 is used for regularization between fully connected layers to suppress overfitting; The output node corresponds to the health status.

[0008] Furthermore, in the full lifecycle status management module, the update method includes an automatic triggering mechanism, which triggers the lifecycle phase to "maintaining" when the health status is faulty.

[0009] Furthermore, the full lifecycle status management module also includes: automatically recording the maintenance records and fault records of the device, wherein the maintenance records include the date and content, and the fault records include the type and fault description.

[0010] Furthermore, in the visualization interaction and decision support module, the method for displaying the real-time status of the device is as follows: the device's current health status, key parameters, and life cycle stage are displayed through the device's monitoring interface.

[0011] Furthermore, in the visualization interaction and decision support module, the method for visually analyzing the real-time status is to draw a multi-parameter time-series trend chart and a health index change chart.

[0012] Furthermore, the method for generating maintenance decisions is as follows: Based on the prediction results, the recommended maintenance time for the health status is calculated as follows: , Calculate the health index corresponding to the stated health status: , Predicted status codes: Normal = 0, Caution = 1, Warning = 2, Alarm = 3, Fault = 4.

[0013] This invention also proposes a method for full lifecycle health management of radar security equipment. The method is implemented based on the full lifecycle health management system for radar security equipment and includes: S1. Initialize system configuration; S2. Collect and record the equipment's full lifecycle data; the full lifecycle data includes equipment operation phase data and process record data throughout each stage of the entire lifecycle; S3. Preprocess the full lifecycle data to construct a time-series feature sequence, and perform data association on the full lifecycle data based on timestamps; S4. Use a CNN time-series prediction model to predict the health status of the time-series feature sequence, obtain the probability distribution of the health status, and take the health status with the highest probability distribution as the prediction result; S5. Perform a status assessment based on the prediction results; S6. Generate maintenance recommendations based on the prediction results; S7. Update the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning; S8. Perform data closed-loop optimization on the full lifecycle data, the prediction results, and the lifecycle stages to update the digital thread data chain; S9. Repeat steps S2 to S8.

[0014] Compared with the prior art, the beneficial effects of the present invention [1] are: This invention discloses a full lifecycle health management system for radar security equipment. It collects data from the equipment's operational phases and process records, creating a complete data record throughout the equipment's lifecycle. A digital thread data processing module preprocesses the lifecycle data and establishes a data chain based on timestamps, enabling data to be correlated chronologically and traceable. A health prediction module uses a three-layer one-dimensional convolutional neural network to predict the health status of time-series feature sequences, outputting health states such as normal, alert, warning, alarm, and fault. A full lifecycle status management module adjusts the lifecycle stages based on the prediction results, updating lifecycle stage information as health status changes. A visualization interaction and decision support module displays real-time status based on the digital thread data chain and plots multi-parameter time-series trend graphs and health index change graphs, displaying real-time information and historical changes of the equipment in trend form and generating maintenance decisions based on the prediction results.

[0015] The present invention discloses a full lifecycle health management method for radar security equipment. By sequentially executing data acquisition, data preprocessing, time-series feature sequence construction, health status prediction, status assessment, maintenance suggestion generation, lifecycle update, and closed-loop optimization, the full lifecycle data, prediction results, and lifecycle stages are continuously updated in the data chain, enabling the digital thread data chain to be continuously supplemented and recycled as the system runs.

[0016] The present invention uses digital thread to drive the formation of a closed loop of full lifecycle data. By associating full lifecycle data according to timestamps, data from different stages are arranged in chronological order to form a digital thread data chain. Through closed-loop optimization steps, full lifecycle data, prediction results, and lifecycle stages are written into the digital thread data chain, so that new associated data records are continuously added to the data chain during the cyclic execution process, thereby achieving data traceability.

[0017] The CNN time-series prediction model described in this invention inputs a time-series feature sequence consisting of N consecutive collected samples into a three-layer one-dimensional convolutional neural network. After processing by convolution, pooling, and dropout, it outputs a probability distribution corresponding to the health status, so that the time-series feature sequence can be directly mapped to a prediction result of one of the following: normal, attention, warning, alarm, or fault.

[0018] This invention employs a multi-dimensional dynamic maintenance decision-making algorithm. By calculating a health index based on predicted status codes, different health states can be converted into corresponding numerical forms. By calculating suggested maintenance times based on the health index, the suggested maintenance times can be automatically updated as the prediction results change. Through a visualization module, multi-parameter time-series trend charts and health index change charts are drawn, so that the changes in time-series parameters and health indices are displayed in the form of trend curves.

[0019] This invention belongs to the field of equipment monitoring technology, and is specifically applied to the health monitoring and maintenance decision-making of radar security equipment. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of a full life-cycle health management system for radar security equipment as described in this invention.

[0022] Figure 2 This is a schematic diagram of the CNN time-series prediction model described in this invention.

[0023] Figure 3 This is a schematic diagram of the real-time status display of the device based on the digital thread data link of the visualization interaction and decision support module described in this invention.

[0024] Figure 4 This is a schematic diagram of a full life-cycle health management method for radar security equipment as described in this invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Specific Implementation Method 1: The radar security equipment full lifecycle health management system described in this implementation method includes: Data acquisition module: used to collect and record the entire lifecycle data of the device; the entire lifecycle data includes data of the device operation phase, as well as process record data throughout each stage of the entire lifecycle; The digital thread data processing module is used to preprocess the full lifecycle data to construct a time-series feature sequence; it is also used to perform data association based on timestamps to obtain a digital thread data chain; the digital thread data chain is traceable and used to backtrack and analyze historical data. Health prediction module: Equipped with a CNN time-series prediction model, the health status of the time-series feature sequence is predicted through a three-layer one-dimensional convolutional neural network of the model, the probability distribution of the health status is obtained, and the health status with the highest probability distribution is taken as the prediction result; the health status includes normal, attention, warning, alarm and fault. The full lifecycle status management module adjusts the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning. Visualization and Decision Support Module: Based on the digital thread data chain, it displays the real-time status of the device, performs visual analysis of the real-time status, and generates maintenance decisions.

[0027] In the data acquisition module, the operational phase data includes the radar equipment's intensity, temperature, humidity, vibration value, and operating current; the process recording data includes equipment manufacturing information, installation date, maintenance records, fault records, and life cycle stage change records.

[0028] The method for constructing the time-series feature sequence in the digital thread data processing module is as follows: Normalization is performed using standardization methods: , Where μ is the mean of the parameters, The standard deviation of the parameter; Extract N consecutive collected samples and construct the time-series feature sequence as follows: , in, is the multi-parameter feature vector at the current time; N=20, which is the sequence length.

[0029] The CNN time-series prediction model is as follows: The model input is the time-series feature sequence, wherein the feature parameters include signal strength, temperature, humidity, vibration, and current; Three one-dimensional convolutional layers: the number of channels is 5, 32, 64 and 128 respectively, the kernel size of each layer is 3, and the padding is 1; Each one-dimensional convolutional layer is followed by a max pooling layer, which has a pooling kernel size of 2 and a stride of 2, and is used for feature dimensionality reduction. The activation function used is the ReLU function; Dropout with a probability of 0.5 is used for regularization between fully connected layers to suppress overfitting; The output node corresponds to the health status.

[0030] In the full lifecycle status management module, the update method includes an automatic triggering mechanism, which triggers the lifecycle phase to "maintaining" when the health status is faulty.

[0031] The full lifecycle status management module also includes: automatically recording the maintenance records and fault records of the device, wherein the maintenance records include the date and content, and the fault records include the type and fault description.

[0032] In the visualization interaction and decision support module, the method for displaying the real-time status of the device is as follows: the device's current health status, key parameters, and life cycle stage are displayed through the device's monitoring interface.

[0033] In the visualization interaction and decision support module, the method for visualizing and analyzing the real-time status is to draw multi-parameter time-series trend charts and health index change charts.

[0034] The method for generating maintenance decisions is as follows: Based on the prediction results, the recommended maintenance time for the health status is calculated as follows: , Calculate the health index corresponding to the stated health status: , Predicted status codes: Normal = 0, Caution = 1, Warning = 2, Alarm = 3, Fault = 4.

[0035] In this embodiment, such as Figure 1 As shown, it includes five core modules: device lifecycle data acquisition module, digital thread data processing module, CNN time series health prediction module, lifecycle status management module, and visualization interaction and decision support module.

[0036] Data flow: The acquisition module provides raw data → the processing module generates time-series features → the prediction module outputs health status → the management module updates the lifecycle and generates maintenance suggestions → the visualization module displays the results and receives operation instructions; Core function: Connecting various modules through digital threads to achieve closed-loop operation of full life cycle health management.

[0037] This system is built upon digital thread data closed-loop control, CNN time series prediction, and full lifecycle state management, encompassing five core modules. The specific system architecture is as follows: (1) Equipment lifecycle data acquisition module This enables comprehensive data collection across all stages of the entire lifecycle, with core functionalities including: Data acquisition during operation: Real-time acquisition of key parameters of radar equipment such as signal strength (S), temperature (T), humidity (H), vibration value (V), and operating current (I). The acquisition interval can be configured to 2 seconds / time by default. Full lifecycle stage data recording: Recording equipment manufacturing information; Manufacturing information includes model and manufacturer; installation date; maintenance records include maintenance date, content, and technician; fault records include fault date, type, and description; and lifecycle stage change records. Data preprocessing: The range of collected data is limited, including signal strength 0-100, temperature -30℃-100℃, and missing values ​​are filled using the mean imputation method to ensure data validity.

[0038] (2) Digital thread data processing module It constructs a closed-loop data acquisition-analysis-prediction-maintenance system, with core functionalities including: Data standardization: Multiple parameters are normalized using a standardization formula, which is: , Where μ is the mean of the parameters, The standard deviation of the parameter; Temporal feature sequence construction: Extract N consecutive collected samples (N is the sequence length, default 20) to construct a temporal feature sequence, using the following formula: , in, is the multi-parameter feature vector at the current time; N=20, which is the sequence length.

[0039] Data storage and traceability: Data at each stage of the entire lifecycle is linked by timestamp to form a traceable digital thread data chain, supporting historical data backtracking and analysis.

[0040] (3) CNN time series health prediction module like Figure 2 As shown, in the CNN time series prediction model structure, the model input is an N×5 dimensional time series (N is the sequence length) → the time series features are extracted through 3 layers of Conv1d+MaxPool1d → after flattening, it is input into 3 fully connected layers → the output is the probability distribution of 5 health states.

[0041] Key nodes: Convolutional layers extract temporal features, Dropout layers prevent overfitting, and fully connected layers complete classification mapping.

[0042] Specifically, the core design for accurate prediction of health status based on multi-parameter time-series features includes: Model input: Time series; The dimension is N×5, where N is the sequence length and 5 represents the five parameters: signal strength, temperature, humidity, vibration, and current. Model structure: The convolutional layers are 3 layers of 1D convolutions (Conv1d), with the input channels being 5→32 channels (kernel_size=3, padding=1), 32→64 channels (kernel_size=3, padding=1), and 64→128 channels (kernel_size=3, padding=1). Pooling layers: Each convolutional layer is followed by a MaxPool1d (kernel_size=2, stride=2) to achieve feature dimensionality reduction; Fully connected layer: After flattening the features, the layers are processed through a 128-dimensional → 64-dimensional → 5-dimensional fully connected layer to output the probability distribution of 5 health states; Activation and regularization: The ReLU activation function is used, and Dropout (probability 0.5) is added between fully connected layers to prevent overfitting; State mapping: The state with the highest output probability is used as the prediction result. The mapping relationship is: 0 - Normal, 1 - Attention, 2 - Warning, 3 - Alarm, 4 - Fault.

[0043] (4) Full lifecycle status management module The full lifecycle health management process is as follows: initialization → data collection and preprocessing → time series construction → CNN model prediction → health status assessment → maintenance suggestion generation → lifecycle update → data closed-loop optimization → repetition.

[0044] Core logic: Using digital threads as a link, dynamic iteration of health management is achieved.

[0045] It enables dynamic management of equipment lifecycle stages and links them with health status. Core functions include: Lifecycle phase division: Define 6 phases (manufacturing, installation, operation, maintenance, replacement, and decommissioned), supporting manual phase updates and automatic triggering (such as automatically switching to "maintenance" in case of failure); Health status is linked to lifecycle: Based on CNN prediction results, the lifecycle stage is dynamically adjusted (e.g., if the prediction is "failure", the stage is automatically updated to "maintaining"). Record Management: Automatically records maintenance records (including date, content, and technician) and fault records (including date, type, and description), and supports traceability by timestamp.

[0046] (5) Visual interaction and decision support module Provides an intuitive monitoring interface and decision support, such as Figure 3 As shown, the visualization interaction and decision support module of this invention displays the real-time status of devices based on a digital thread data chain, and provides examples of security coverage maps and attention heatmaps. It includes a device status display area (health index, key parameters, life cycle stage), a trend map area (multi-parameter time-series trends, health index changes), a maintenance suggestion area (suggested maintenance time and content), and a control operation area (start / stop monitoring, model update).

[0047] Core features include: Multidimensional dynamic maintenance decision algorithm [2]: integrates health index, failure probability and life cycle stage to generate personalized maintenance suggestions.

[0048] Real-time status display: Shows the device's current health status, key parameter values, and lifecycle stage; Trend visualization: Draws multi-parameter time-series trend charts and health index change charts, supporting visualization analysis of the most recent 50-100 data points; Maintenance decision generation: Based on the predicted health status, the recommended maintenance time is calculated using the following formula: , Health Index Calculation: This quantifies health status based on predicted condition. The formula is as follows: , Predicted status codes: Normal = 0, Caution = 1, Warning = 2, Alarm = 3, Fault = 4.

[0049] Specific implementation method two, such as Figure 4 As shown in this embodiment, a method for full lifecycle health management of radar security equipment is implemented based on the radar security equipment full lifecycle health management system. The method includes: S1. Initialize system configuration; S2. Collect and record the equipment's full lifecycle data; the full lifecycle data includes equipment operation phase data and process record data throughout each stage of the entire lifecycle; S3. Preprocess the full lifecycle data to construct a time-series feature sequence, and perform data association on the full lifecycle data based on timestamps; S4. Use a CNN time-series prediction model to predict the health status of the time-series feature sequence, obtain the probability distribution of the health status, and take the health status with the highest probability distribution as the prediction result; S5. Perform a status assessment based on the prediction results; S6. Generate maintenance recommendations based on the prediction results; S7. Update the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning; S8. Perform data closed-loop optimization on the full lifecycle data, the prediction results, and the lifecycle stages to update the digital thread data chain; S9. Repeat steps S2 to S8.

[0050] In this embodiment, the health management process described in this invention has a logical closed loop, executing cyclically according to "initialization → data closed-loop operation → predictive decision-making → maintenance and optimization". The specific steps are as follows: S1: System Initialization (Input: Basic device information; Output: Initial configuration) Configure basic information such as device ID, model, and installation date; Set the acquisition parameters (interval 2 seconds, sequence length N=20) and CNN training parameters (epochs=50, batch_size=32, learning rate 0.001); Initialize the lifecycle phase (installing / running) and the digital thread data chain.

[0051] S2: Data Acquisition and Preprocessing (Input: Real-time equipment data; Output: Standardized time series) Simultaneously collect 5 types of core parameters and record lifecycle events (maintenance / fault); Scope limitation, missing value imputation, and standardization; Construct a time-series feature sequence of length N to provide input for the model.

[0052] S3: CNN Model Training and Deployment (Input: Time Series + Historical Labels; Output: Trained Model) Generate a training set (≥1000 samples, labeled with actual health status); Using the MSELoss loss function and the Adam optimizer, train for 50 epochs until the test accuracy is ≥85%; Deploy the model to the system for real-time prediction.

[0053] S4: Health Status Prediction (Input: Standardized time series; Output: Predicted status + Health index) The model takes a preprocessed time series as input and outputs a probability distribution of 5 states. Calculate the health index and failure probability (failure probability = status code × 25%). Combine the results of lifecycle phase corrections (such as reducing the failure probability weight of equipment during maintenance).

[0054] S5: Dynamic Maintenance and Phased Updates (Input: Prediction Results; Output: Maintenance Suggestions + Phased Updates) Based on predicted status and health index, maintenance time and content are generated; Abnormal conditions (alarms / faults) automatically switch lifecycle stages and trigger warnings; Record and maintain execution results, and update the digital thread data chain.

[0055] S6: Closed-loop optimization (Input: Newly collected data + Maintenance records; Output: Optimized model + Data chain) Iteratively update the training dataset (add new data to the sample library); The model is retrained every 5 maintenance cycles to improve prediction accuracy; Visualize the optimization effects and health trends.

[0056] Specific implementation method three: This implementation method is for large-scale security scenarios and uses the radar security equipment full life cycle health management method described in this invention to carry out full-process radar equipment health management.

[0057] (a) Environmental configuration Hardware environment Computing devices: GPUs that support CUDA (such as NVIDIA RTX 3060 and above) or CPUs (suitable for small-scale deployments); Radar equipment: Security radar that supports output of signal strength, temperature, humidity, vibration, and current parameters (coverage radius ≥200m); Storage and Network: ≥10GB of free space (for storing data, models, and logs), LAN bandwidth ≥100Mbps (to ensure real-time data transmission).

[0058] Software environment Basic framework: Python 3.8-3.10, PyTorch 2.0+ (with CUDA 11.7+ for model training); Data processing and visualization: Pandas 1.5+, NumPy 1.24+, Matplotlib 3.7+, PyQt5 5.15+ (GUI interface); Auxiliary tools: Scikit-learn 1.2+ (data standardization, dataset splitting).

[0059] (II) Implementation Parameter Configuration Equipment parameters: 2 radar devices (ID: RAD-2023-001, RAD-2023-002), model SR-1000 / SR-2000, installed on 2025-06-15 and 2025-03-22 respectively; Acquisition parameters: Acquisition interval 2 seconds, sequence length N=20, parameter range (signal strength 0-100, temperature -30℃-100℃, vibration 0-2.0, current 0-5.0A); Model parameters: Conv1d channel count 32→64→128, fully connected layer dimension 128→64→5, epochs=50, batch_size=32, learning rate=0.001; Maintenance parameters: A health index threshold of ≥80 is considered normal, while ≤60 triggers an alert.

[0060] (III) Implementation Steps S1: Initialization Enter the basic information of the device and set the initial stage of its lifecycle to "running". Configure the data acquisition and model parameters to complete system deployment.

[0061] S2: Data Acquisition and Preprocessing Real-time acquisition of signal strength, temperature, humidity, vibration, and current data from two devices; For outlier data (such as temperature > 100℃), range correction is performed, missing values ​​are filled with the mean, and time series are constructed after standardization.

[0062] S3: Model Training and Deployment 1200 training samples were generated based on historical data, and the CNN model was trained to a test accuracy of 88% before being deployed to the system. S4: Health Prediction and Status Assessment Inputting the time series sequence into the model, RAD-2023-001 predicts the state as "normal" (state code 0), with a health index of 100 and a failure probability of 0%; RAD-2023-002 predicts the state as "attention" (state code 1), with a health index of 80 and a failure probability of 25%. S5: Maintenance Recommendation Generation RAD-2023-001 generated a maintenance recommendation: "Equipment is in good condition; routine inspection is recommended after 45 days." RAD-2023-002 generated a maintenance recommendation: "The equipment is experiencing a minor abnormality. It is recommended to perform vibration parameter calibration in 20 days." S6: Closed-loop optimization Record the operating data and maintenance plan execution status of the two devices; Thirty days later, based on 100 newly collected data points, the model was iteratively trained, and the accuracy was improved to 90%.

Claims

1. A full lifecycle health management system for radar security equipment, characterized in that, The system includes: Data acquisition module: used to collect and record the entire lifecycle data of the device; the entire lifecycle data includes data of the device operation phase, as well as process record data throughout each stage of the entire lifecycle; The digital thread data processing module is used to preprocess the full lifecycle data to construct a time-series feature sequence; it is also used to perform data association based on timestamps to obtain a digital thread data chain; the digital thread data chain is traceable and used to backtrack and analyze historical data. Health prediction module: Equipped with a CNN time-series prediction model, the health status of the time-series feature sequence is predicted through a three-layer one-dimensional convolutional neural network of the model, the probability distribution of the health status is obtained, and the health status with the highest probability distribution is taken as the prediction result; the health status includes normal, attention, warning, alarm and fault. The full lifecycle status management module adjusts the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning. Visualization and Decision Support Module: Based on the digital thread data chain, it displays the real-time status of the device, performs visual analysis of the real-time status, and generates maintenance decisions.

2. The radar security equipment full life cycle health management system according to claim 1, characterized in that, In the data acquisition module, the operational phase data includes the radar equipment's intensity, temperature, humidity, vibration value, and operating current; the process recording data includes equipment manufacturing information, installation date, maintenance records, fault records, and life cycle stage change records.

3. The radar security equipment full life cycle health management system according to claim 1, characterized in that, The method for constructing the time-series feature sequence in the digital thread data processing module is as follows: Normalization is performed using standardization methods: ， Where μ is the mean of the parameters, The standard deviation of the parameter; Extract N consecutive collected samples and construct the time-series feature sequence as follows: ， in, is the multi-parameter feature vector at the current time; N=20, which is the sequence length.

4. The radar security equipment full life cycle health management system according to claim 1, characterized in that, The CNN time-series prediction model is as follows: The model input is the time-series feature sequence, wherein the feature parameters include signal strength, temperature, humidity, vibration, and current; Three one-dimensional convolutional layers: the number of channels is 5, 32, 64 and 128 respectively, the kernel size of each layer is 3, and the padding is 1; Each one-dimensional convolutional layer is followed by a max pooling layer, which has a pooling kernel size of 2 and a stride of 2, and is used for feature dimensionality reduction. The activation function used is the ReLU function; Dropout with a probability of 0.5 is used for regularization between fully connected layers to suppress overfitting; The output node corresponds to the health status.

5. The radar security equipment full life cycle health management system according to claim 1, characterized in that, In the full lifecycle status management module, the update method includes an automatic triggering mechanism, which triggers the lifecycle phase to "maintaining" when the health status is faulty.

6. The radar security equipment full life cycle health management system according to claim 1, characterized in that, The full lifecycle status management module also includes: automatically recording the maintenance records and fault records of the device, wherein the maintenance records include the date and content, and the fault records include the type and fault description.

7. The radar security equipment full life cycle health management system according to claim 1, characterized in that, In the visualization interaction and decision support module, the method for displaying the real-time status of the device is as follows: the device's current health status, key parameters, and life cycle stage are displayed through the device's monitoring interface.

8. The radar security equipment full life cycle health management system according to claim 1, characterized in that, In the visualization interaction and decision support module, the method for visualizing and analyzing the real-time status is to draw multi-parameter time-series trend charts and health index change charts.

9. The radar security equipment full life cycle health management system according to claim 1, characterized in that, The method for generating maintenance decisions is as follows: Based on the prediction results, the recommended maintenance time for the health status is calculated as follows: ， Calculate the health index corresponding to the stated health status: ， Predicted status codes: Normal = 0, Caution = 1, Warning = 2, Alarm = 3, Fault = 4.

10. A method for full lifecycle health management of radar security equipment, characterized in that, The method is implemented based on the full lifecycle health management system for radar security equipment as described in any one of claims 1 to 9, and the method includes: S1. Initialize system configuration; S2. Collect and record the equipment's full lifecycle data; the full lifecycle data includes equipment operation phase data and process record data throughout each stage of the entire lifecycle; S3. Preprocess the full lifecycle data to construct a time-series feature sequence, and perform data association on the full lifecycle data based on timestamps; S4. Use a CNN time-series prediction model to predict the health status of the time-series feature sequence, obtain the probability distribution of the health status, and take the health status with the highest probability distribution as the prediction result; S5. Perform a status assessment based on the prediction results; S6. Generate maintenance recommendations based on the prediction results; S7. Update the lifecycle stage of the equipment based on the prediction results; the lifecycle stage includes manufacturing, installation, operation, maintenance, replacement, and decommissioning; S8. Perform data closed-loop optimization on the full lifecycle data, the prediction results, and the lifecycle stages to update the digital thread data chain; S9. Repeat steps S2 to S8.

Images