Method and system for monitoring full life cycle of leasing equipment based on Internet of Things

Abstract

The invention provides a leasing equipment full life cycle monitoring method and system based on the Internet of Things, and belongs to the technical field of equipment operation and maintenance monitoring, and the method comprises the steps: obtaining real-time state data of leasing equipment in an operation process through a multi-source sensor; performing feature extraction on the real-time state data to obtain various feature vectors, and constructing a multi-dimensional feature vector set; inputting the multi-dimensional feature vector set into a random forest model, and judging whether the leasing equipment is in an abnormal state or not in combination with a space-time clustering algorithm; if yes, fault prediction is carried out on the leasing equipment in combination with an abnormal detection result, historical operation data and a baseline model, and a fault probability value of the leasing equipment is obtained; the future position of the leasing equipment is predicted based on the position data, an electronic fence boundary is set for the leasing equipment, if the future position exceeds the electronic fence boundary, an early warning notification of trajectory deviation is generated, and the system finally realizes trajectory and fault visual presentation in a thermodynamic diagram mode, so that the equipment operation and maintenance intelligence and management precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment operation and maintenance monitoring, and specifically to a full life cycle monitoring method and system for rental equipment based on the Internet of Things, which belongs to the technical scope of industrial Internet of Things and intelligent fault detection and trajectory management systems. Background Art

[0002] With the advancement of industrial digitalization and intelligent operations and maintenance, the demand for remote monitoring of leased assets such as construction machinery, energy equipment, and logistics equipment is growing. Traditional equipment status monitoring methods often rely on fixed-point detection of a single physical quantity (such as temperature, vibration, or pressure). These methods fail to fully reflect the evolving trends of the equipment's operating status and are prone to false positives or omissions. In recent years, the application of the Internet of Things and edge computing has driven the development of multi-source heterogeneous sensor technology, enabling the continuous collection of equipment status data across time and space. However, the fusion and analysis of multi-dimensional time series data, such as vibration, temperature, pressure, and position, remains a core technical challenge in current engineering practice.

[0003] In terms of anomaly detection, traditional methods rely on static threshold settings, which are difficult to adapt to different equipment types and dynamic working conditions, and cannot adjust the judgment benchmark in real time. In terms of fault prediction, most existing technologies use a single model to perform trend regression on historical data, ignoring the long-term and short-term dependencies of status data, resulting in insufficient prediction accuracy. In terms of trajectory management, static electronic fences are often used for out-of-bounds judgment, but this method cannot dynamically reflect the adjustment of the operation boundary, nor can it identify the inaccessible areas in the operation area, which poses a risk of misjudgment. In addition, existing monitoring systems generally lack the ability to express the spatial visualization of early warning results, making it difficult for operation and maintenance personnel to perceive abnormal equipment areas and fault trends in a timely manner.

[0004] Therefore, how to construct a monitoring method that combines multi-source data fusion, dynamic anomaly recognition, intelligent fault prediction and spatial deviation judgment for rental equipment application scenarios has become the technical problem to be solved by the present invention. Summary of the Invention

[0005] The present invention provides a full life cycle monitoring method and system for rental equipment based on the Internet of Things, aiming to solve technical problems existing in the existing technology, such as the difficulty in fusing and analyzing multi-source sensor data, static anomaly detection, inaccurate fault prediction, distorted trajectory deviation determination, and insufficient visualization of results.

[0006] In a first aspect, the present application provides a method for monitoring the entire life cycle of rental equipment based on the Internet of Things, the method comprising:

[0007] S1. Monitor the operation process of the rental equipment through multi-source sensors to obtain real-time status data, including position data, vibration data, pressure data, and temperature data;

[0008] S2. Extract features from the real-time status data to obtain multiple feature vectors of the rental equipment and construct a multi-dimensional feature vector set;

[0009] S3. Inputting the multidimensional feature vector set into a random forest model and combining it with a spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state;

[0010] S4. If yes, then combine the abnormality detection results, historical operation data, and the preset baseline model to predict the failure of the leased equipment and obtain the failure probability value of the leased equipment;

[0011] S5. Predict the future location of the rental device based on the location data, set an electronic fence boundary for the rental device, and generate a trajectory deviation warning notification if the future location exceeds the electronic fence boundary, and push the warning information to the management platform.

[0012] Furthermore, feature extraction is performed on the real-time status data, including:

[0013] Feature extraction is performed on the real-time status data based on a convolutional neural network model. The convolutional neural network model includes a deep feature fusion layer composed of multiple convolution and pooling structures, which is used to extract the frequency pattern of the vibration data, the mutation pattern of the pressure data, and the periodic pattern of the temperature data.

[0014] Furthermore, before inputting the multidimensional feature vector set into the random forest model, the following steps are performed:

[0015] S31: Continuously sampling each type of feature data in the multidimensional feature vector set based on a sliding time window;

[0016] S32: Calculate the mean and standard deviation of each type of feature data within a preset window to construct a dynamic statistical feature set;

[0017] S33: When the deviation of the newly collected feature data from the mean value in the corresponding window exceeds a predetermined abnormal trigger threshold, the feature data is determined to be an abnormal warning point;

[0018] S34: The abnormal warning result is used as the prior judgment of the random forest model as input.

[0019] Furthermore, determining whether the rental equipment is in an abnormal state includes:

[0020] S41: inputting the multidimensional feature vector set into a trained random forest model, wherein the trained random forest model is trained by a training data set including historical operation data and corresponding status labels of the rental equipment at each usage stage;

[0021] S42: The random forest model performs preliminary classification processing based on the input feature data and outputs the predicted probability of each state category;

[0022] S43: Marking the data at the time point where the abnormality occurs based on the predicted probability;

[0023] S44: Based on the spatiotemporal clustering algorithm, perform clustering analysis on the data points marked as abnormal in the spatial and temporal dimensions to identify whether there are group, regional or stage-specific abnormal behaviors and obtain clustering results;

[0024] S45: Based on the clustering results, forming an abnormal pattern of a spatially continuous area or time period;

[0025] S46: Compare the abnormal pattern with the common behavior pattern of the rental equipment in the historical data, calculate the similarity, and mark it as an abnormal state if the similarity is greater than a preset first threshold.

[0026] Furthermore, the baseline model records the health parameter intervals of the rental equipment at different operating stages, which are used to identify trend changes in the time series.

[0027] Furthermore, fault prediction is performed on rental equipment, including:

[0028] S61: Use the anomaly detection results as explicit state variables in the input features of the LSTM model to identify short-term operational anomalies of the equipment.

[0029] S62: Slice the historical operating data of the rental equipment using fixed time windows to construct a multi-dimensional time series input matrix, where each row in the matrix corresponds to a time segment and each column corresponds to a different perception feature or state label;

[0030] S63: Inputting the time series input matrix into an LSTM model, wherein the LSTM model extracts dynamic change characteristics of the rental equipment during long-term operation;

[0031] S64: The comprehensive feature vector formed by combining the anomaly detection results, the dynamic change characteristics of historical operation data, and the baseline model is input into the LSTM model. Through its gating mechanism and memory function, the LSTM automatically captures abnormal trends and potential changes in the long-term operation of the equipment, thereby generating a hidden state vector.

[0032] S65: Based on the output of the LSTM model, the dynamic change characteristics of the leased equipment are compared with the healthy parameter range in the baseline model to calculate the failure risk of the equipment during the forecast period.

[0033] Furthermore, the electronic fence boundary is composed of polygonal geometric constraints dynamically generated by a cloud server. The dynamically generated polygonal geometric constraints are semantic fences constructed based on high-precision map data and can be adjusted in real time according to the actual operation area.

[0034] Furthermore, the semantic fence built based on high-precision map data includes:

[0035] S81: Constructing a rough boundary region based on the location data and historical operation data of the rental equipment;

[0036] S82: Extracting, within the rough boundary area, multidimensional spatial constraint factors within the area based on elevation data and image data of the working area where the rental equipment is located, the multidimensional spatial constraint factors including elevation gradient, slope, and obstruction information;

[0037] S83: performing pixel-level classification and identification on the multidimensional spatial constraint factors, identifying inaccessible areas for rental equipment and obstacle information, and marking them as restricted areas as spatial boundary constraint conditions of the semantic fence;

[0038] S84: Analyze the boundary information of the obstacle data and generate a restriction polygon that allows the rental equipment to move based on the historical operation behavior of the rental equipment;

[0039] S85: The semantic fence supports a nested structure, and multiple restricted areas are integrated into the boundary of the restricted polygon to form the nested structure, thereby obtaining the semantic fence.

[0040] Furthermore, after S5, it includes:

[0041] S6. Based on the trajectory deviation data in the warning notification, the degree of deviation between the trajectory point and the electronic fence boundary is calculated using Euclidean distance, and the trajectory deviation distribution and fault probability distribution are displayed on a visual interface using heat map rendering technology to achieve visual monitoring of the entire life cycle of the rental equipment;

[0042] S7. Generate a trajectory deviation distribution map and a failure probability distribution map of the rental equipment through heat map rendering to assist the management platform in visual analysis.

[0043] In a second aspect, the present application provides a full life cycle monitoring system for rental equipment based on the Internet of Things, the system comprising:

[0044] An acquisition module is used to monitor the operation process of the rental equipment through multi-source sensors and obtain real-time status data, which includes position data, vibration data, pressure data, and temperature data;

[0045] An extraction module, configured to extract features from the real-time status data, obtain multiple feature vectors of the rental equipment, and construct a multi-dimensional feature vector set;

[0046] An abnormality judgment module, used to input the multidimensional feature vector set into a random forest model and combine it with a spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state;

[0047] The fault diagnosis module is used to determine if the fault is true, and then combine the abnormal detection results, historical operation data and the preset baseline model to predict the fault of the rental equipment and obtain the fault probability value of the rental equipment;

[0048] A location prediction module is used to predict the future location of the rental equipment based on the location data, set an electronic fence boundary for the rental equipment, and generate a trajectory deviation warning notification if the future location exceeds the electronic fence boundary, and push the warning information to the management platform.

[0049] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0050] 1. Achieve structured fusion and deep feature extraction of multi-source heterogeneous operation data, effectively improving the expressiveness and information utilization of equipment status data;

[0051] 2. A dynamic threshold mechanism based on the statistical characteristics of a sliding window enhances the adaptability and sensitivity of abnormality identification to working conditions, avoiding misjudgments caused by fixed thresholds.

[0052] 3. By introducing spatiotemporal features to participate in classification and clustering, the temporal continuity and spatial consistency of anomaly detection can be achieved, and the model's ability to understand the equipment operation mode can be enhanced;

[0053] 4. Use long-short-term memory networks to model the dependency of state sequences, improving the temporal correlation and prediction accuracy of fault probability prediction;

[0054] 5. Kalman filtering and dynamic semantic fencing are used to predict trajectory deviations based on spatial topological relationships, effectively avoiding high-risk behaviors such as equipment crossing boundaries and entering obstacle areas.

[0055] 6. The degree of trajectory deviation is quantified with the help of Euclidean distance, and dynamic rendering is performed through heat maps to improve the intuitiveness and response efficiency of operation and maintenance personnel in identifying warning areas and risk distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 This is a schematic diagram of an embodiment of a method for monitoring the entire life cycle of rental equipment based on the Internet of Things in an embodiment of the present application;

[0058] Figure 2 This is a process diagram of the rental equipment from data collection to early warning notification in an embodiment of the present application;

[0059] Figure 3 A comprehensive schematic diagram of abnormal identification of rental equipment in an embodiment of the present application;

[0060] Figure 4 This is a performance comparison chart of the fault prediction model in the embodiment of this application;

[0061] Figure 5 This is a schematic diagram of an embodiment of the Internet of Things-based rental equipment full life cycle monitoring system in the embodiment of the present application. DETAILED DESCRIPTION

[0062] The embodiments of the present application provide a method and system for monitoring the entire life cycle of rental equipment based on the Internet of Things. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0063] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for monitoring the entire life cycle of rental equipment based on the Internet of Things includes:

[0064] Step S1: Monitor the operation process of the rental equipment through multi-source sensors to obtain real-time status data, which includes position data, vibration data, pressure data and temperature data.

[0065] Specifically, in traditional rental equipment management, real-time monitoring and fault warnings rely on manual inspections or single sensor data. This often prevents timely detection of equipment failures, impacting equipment efficiency and maintenance costs. With the development of IoT technology, it has become possible to leverage multi-source sensor data (such as vibration, temperature, and pressure) for real-time equipment monitoring. However, existing technologies often struggle to comprehensively analyze historical equipment operating data, environmental data, and real-time status data. They lack effective algorithmic models for accurately predicting equipment failures, and are unable to perform real-time predictions of equipment locations and intelligently manage equipment fencing. Therefore, this application proposes a method for monitoring rental equipment throughout its lifecycle and providing timely fault warnings based on multi-source data.

[0066] First of all, the leased equipment in this application includes but is not limited to construction equipment (such as tower cranes, excavators, concrete mixers, etc.), agricultural equipment (such as tractors, harvesters, etc.), transportation equipment (such as forklifts, stackers, etc.) and energy equipment (such as generator sets, pumping stations, etc.).

[0067] The following takes heavy-duty excavation equipment in an engineering operation scenario as an example. During the operation of the excavation equipment, real-time status data is collected through multi-source sensor modules deployed on the excavation equipment. Vibration data is obtained through accelerometers or vibration sensors to monitor the vibration status of the tower crane; temperature data is collected by thermistor sensors at a set sampling frequency to continuously collect time-series fluctuation values, indicating the thermal change trend of the equipment environment or core components; pressure data is collected through pressure sensors, indicating the dynamic pressure value inside the hydraulic system or pneumatic system, as an indicator reflecting load changes and fluid control status; position data is obtained through GPS sensors or wireless communication modules to obtain the precise position of the excavation equipment in real time, providing dynamic coordinate changes of the excavation equipment in space for subsequent trajectory modeling and spatial boundary judgment.

[0068] The above-mentioned types of data are inconsistent in physical dimensions, time scales and change characteristics. In order to achieve unified processing and data fusion, this application also introduces an edge computing unit as a data processing front end in the data acquisition stage. The edge computing unit is deployed in the equipment field controller or the proximal communication node, has preliminary computing capabilities and is embedded with data preprocessing logic. After the acquisition of various types of raw data, the edge computing module performs corresponding feature preprocessing processes for the signal types corresponding to different data, where vibration data is a frequency domain signal, temperature data is a time series fluctuation signal, pressure data is a dynamic change signal, and position data is a longitude and latitude sequence. Feature indicators such as the main peak amplitude of the spectrum and the energy density in the frequency domain are extracted from the vibration signal; statistical features such as the fluctuation amplitude and the rate of change in the time window are extracted from the temperature and pressure signals; the current coordinate point, the difference before and after, and the trajectory direction change information are extracted from the position signal, and the above processing results are cached in a structured form in the local storage unit at the edge end. After the data is stable, it is sent to the cloud through a low-power wide area network to improve transmission efficiency and real-time guarantee capabilities.

[0069] After edge processing, the original vibration data still has high-frequency noise interference. In order to improve the stability and effectiveness of subsequent feature extraction, the present invention uses a wavelet denoising method to perform multi-scale analysis on the vibration signal, and separates the noise frequency band and the main frequency band signal, and obtains the vibration spectrum result after noise suppression by reconstructing the signal. Subsequently, in order to eliminate the inconsistency of different signal data in dimension, value range and dynamic characteristics, all collected data are subjected to maximum and minimum value normalization processing respectively, so that they are expressed in a unified numerical range to form a standardized sensor data set. In order to ensure the stable upload of edge collected data and the coordinated transmission of data from multiple nodes, the data is sent to the cloud server through the NB-IoT communication module.

[0070] Step S2: extract features from the real-time status data to obtain multiple feature vectors of the rental equipment and construct a multi-dimensional feature vector set.

[0071] Specifically, if Figure 2 As shown, it is a process diagram of rental equipment from data collection to early warning notification. This application uses convolutional neural networks to construct a feature extraction model to achieve unified modeling and significant feature extraction of multi-source data. The convolutional neural network model consists of multiple convolution kernels with shared weights, which act on feature channels of different dimensions respectively. The vibration channel uses small-scale convolution kernels to extract the main components in the frequency domain, the temperature channel uses medium-scale convolution to identify periodic fluctuation segments, and the pressure channel uses a cross-scale convolution structure to identify high-frequency mutations and low-frequency interval drifts. In order to avoid a single channel feature dominating the network learning direction, feature normalization masking and cross-channel attention mechanism are introduced in the multi-channel fusion stage, and the three types of feature vectors such as vibration, temperature, and pressure are uniformly represented in the intermediate fusion layer to construct a multi-dimensional feature vector set as the basic input for subsequent classification and prediction modules.

[0072] This multidimensional feature vector set retains the main dynamic change patterns under different signal dimensions, and expresses the interdependence between features through parameter weights learned by the network. It has temporal correlation, physical logical integrity and spatial coupling structure consistency.

[0073] During the operation of the excavation equipment, the vibration frequency increases, the pressure fluctuates violently, and the temperature rises abnormally. The weights of multiple channel features in the fused feature vector increase together. The model can immediately reflect this state as a potential abnormality, thereby significantly improving the recognition accuracy and response time. It solves the existing problems of incomplete single-channel signal extraction, difficult pattern separation, and delayed system response, and meets the key requirements of complex data fusion modeling and pre-identification of fault warnings in intelligent operation and maintenance scenarios.

[0074] Step S3: Input the multidimensional feature vector set into the random forest model and combine it with the spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state.

[0075] Specifically, if Figure 3 As shown, it is a comprehensive schematic diagram of equipment abnormality identification. Based on the multi-dimensional feature vector set extracted in step S2, abnormality identification processing is performed on the operating status of the rental equipment. First, this application adopts a set threshold trigger mechanism for initial screening. When any type of feature value such as vibration frequency amplitude, temperature fluctuation gradient or pressure change interval exceeds the corresponding preset abnormal threshold range, this type of feature is marked as a suspected abnormal item and enters the status classification judgment process.

[0076] Next, the multidimensional feature vector set is fed into a random forest model to identify abnormal conditions. The random forest model uses a tree-structured ensemble classification algorithm to construct multiple weak classifiers. The model then classifies the status of the rental equipment based on the multidimensional feature vectors, yielding various status categories, including "normal," "mildly abnormal," and "severely abnormal." To improve the adaptability and spatial generalization of the judgment results, the model input also incorporates the time at which the feature data was collected as a priori input for the random model. This will be discussed in detail later.

[0077] In order to enhance the accurate recognition capability of anomaly discrimination results, this application also introduces spatiotemporal clustering algorithms, such as DBSCAN or density neighborhood clustering, to perform spatial distribution density analysis on short-term continuous triggered anomaly labels, identify whether there are group, regional or stage-specific abnormal behaviors, and form an abnormal risk report for geographical areas. The specific implementation process will be explained later.

[0078] For example, in the case of large-scale engineering equipment rentals, if a certain model of equipment in a certain area experiences frequent and sudden pressure changes accompanied by a continuous temperature rise, the system automatically identifies or utilizes a random forest model to identify the equipment as being in an abnormal load state. It then uses a spatial clustering mechanism to determine whether this represents a regional risk, thereby generating proactive operational intervention recommendations. This approach effectively addresses the issues of poor model generalization, inability to identify geographic-related anomalies, and high false alarm rates in traditional anomaly detection. It possesses practicality and deployment efficiency for engineering applications, constituting a key intelligent identification link in the equipment monitoring system of the present invention.

[0079] Step S4: If yes, then combine the abnormality detection results, historical operation data and the preset state baseline model to predict the failure of the rental equipment and obtain the failure probability value of the rental equipment.

[0080] Specifically, to dynamically model the operating status of rental equipment and predict fault risks, the system constructs a sequence feature input set based on identified anomaly detection results, long-term historical operating data, and a preset state baseline model. This system then uses a long-short-term memory network to model the temporal dependencies of this sequence. Anomaly detection results constitute the explicit state variables within the input features, marking short-term operational anomalies. Historical operating data includes raw or extracted features such as vibration, temperature, and pressure at multiple time points. A preset baseline model records the health parameter ranges of the equipment at different operating stages, serving as a comparison scale to assist the network in identifying trend changes in the time series.

[0081] During data preparation, the system slices structured device operation data into fixed time windows to construct a multidimensional time series input matrix, where each row corresponds to a time segment and each column corresponds to a different perceptual feature or state label. This input matrix is ​​then fed into a long short-term memory (LSTM) network. Using a gating mechanism, the system learns the temporal evolution patterns and long-term dependencies between input features, automatically capturing the accumulation of abnormal trends or latent changes during device operation. Because LSTMs can memorize feature changes over long time spans, their output latent state vectors effectively reflect the state transition patterns of devices over continuous operation cycles.

[0082] During the training phase, the system adjusts network parameters based on historically annotated data, enabling the LSTM to learn how different state characteristics contribute to the future health of the device. During the prediction phase, the system leverages current operating data and the LSTM model to generate a trajectory of the device's state evolution over the prediction period and outputs a corresponding failure probability. This probability is normalized to the [0, 1] range using a Sigmoid mapping function to reflect the likelihood of future failure or significant deviation of the device under the current state.

[0083] Throughout the entire process, the risk labels output by the anomaly detection module serve not only as part of the network input, but also as prior knowledge input for the LSTM model's prediction targets, effectively improving the model's accuracy in identifying fault precursors and enhancing the system's ability to respond to weak trend signals. This technical approach tightly couples the algorithm's features with sensor perception, anomaly discrimination, and early warning generation processes, possessing structural integrity and functional synergy. It can effectively address the problem in existing methods where single-point anomalies are difficult to infer as trend failures, and is suitable for reliability prediction and maintenance scheduling of rental equipment in multiple scenarios. It has significant technical advantages in improving equipment safety and operation and maintenance efficiency. As a key component of the overall technical solution, this section embodies the fusion and innovation of algorithmic structure and technical means in the direction of fault prediction, and possesses significant engineering practical value and creativity.

[0084] Step S5: Predict the future location of the rental equipment based on the location data, set an electronic fence boundary for the rental equipment, and generate a trajectory deviation warning notification if the future location exceeds the electronic fence boundary, and push the warning information to the management platform.

[0085] Specifically, to achieve linked perception and risk warning of the operating status and spatial behavior of rental equipment, the system uses a Kalman filter algorithm to continuously and dynamically predict the equipment's future spatial trajectory based on the equipment failure probability value generated in step S4 and the currently collected location information. The Kalman filter uses the current latitude and longitude position, velocity vector, and acceleration estimate as state inputs. It constructs a prediction model using a state transition matrix and iteratively optimizes the model by integrating historical trajectory information with sensor noise covariance information. The resulting predicted trajectory is expressed as a time series, forming a trajectory point set consisting of multiple latitude and longitude coordinates.

[0086] The spatial topological relationship between the trajectory point set and the geo-fence boundary data generated in real time by the cloud server is determined. The geo-fence boundary is a polygonal geometric constraint constructed by a dynamic boundary generation algorithm based on the actual geographic scope of the operation area and task requirements. This boundary can be dynamically adjusted based on changing conditions such as the equipment's operating area and task type. The system sequentially checks the predicted trajectory points against the fence boundary for spatial inclusion. If one or more trajectory points fall outside the boundary, the device is considered to be on a trajectory deviation trend, triggering a trajectory anomaly warning condition.

[0087] At the same time, the system compares the device's current failure probability value with a preset threshold. If the probability value exceeds the set tolerance limit, the device is deemed to have entered a high-risk range for potential failure, constituting a fault-related abnormality warning condition. When any of these warning conditions is triggered, the system automatically generates a warning notification containing multiple structured fields, including the current fault status assessment value, predicted trajectory point offset, offset direction, and corresponding fence boundary point. This warning data packet is then uploaded to the device management platform via the NB-IoT communication module using the low-power IoT protocol. Based on this notification, the platform visualizes the device's current and future operational risks and drives the coordinated execution of corresponding strategies by systems such as scheduling and maintenance.

[0088] The above technical solution forms a complete risk warning and judgment mechanism by integrating fault trend modeling in the time domain with dynamic trajectory prediction in the spatial domain. The Kalman filter model establishes a logical association with the equipment status probability output in the prediction stage. The electronic fence boundary and trajectory offset judgment results support each other and have significant linkage. The algorithm characteristics are combined with physical state detection and map boundary constraints to form a functionally consistent structure. It can effectively solve the problems of fault identification lag and insensitive trajectory offset warning in existing technologies, significantly improve the monitoring system's ability to proactively judge the operating status of equipment in complex operating environments, and is suitable for intelligent management scenarios of mobile rental equipment.

[0089] In a specific embodiment, the feature extraction of the real-time status data in step S2 specifically includes the following steps:

[0090] Feature extraction of real-time status data is performed based on a convolutional neural network model. The convolutional neural network model includes a deep feature fusion layer composed of multiple convolution and pooling structures, which is used to extract the frequency pattern of vibration data, the mutation pattern of pressure data, and the periodic pattern of temperature data.

[0091] Specifically, the present application performs feature extraction on real-time status data based on a convolutional neural network model, wherein the convolutional neural network model includes a deep feature fusion layer composed of multiple convolutional layers and pooling layers. This design can automatically learn and extract key features in the data. The convolution layer extracts local features of the data through local perception areas, such as frequency patterns in vibration data, mutation patterns in pressure data, and periodic patterns in temperature data. Specifically, the convolution operation slides the filter on the input data, thereby gradually extracting features of different scales in the time or space dimension. The parameters of these convolutional layers are optimized through training data, so that the model can gradually learn the characteristics of different signal patterns and retain them in the network for subsequent analysis.

[0092] For vibration data, convolutional neural networks learn frequency patterns to capture periodic vibrations or frequency anomalies that may occur during equipment operation. These anomalies are often associated with equipment failure or overload. For pressure data, convolutional neural networks focus on capturing sudden changes in pressure signals over short periods of time, which typically occur at the moment of equipment failure or anomaly. While extracting these sudden changes through pooling layers, the network also compresses irrelevant data, increasing its sensitivity to key anomaly patterns. For temperature data, convolutional neural networks can extract periodic patterns, namely, the temperature variations of equipment during normal operation. Temperature typically fluctuates with the load on the equipment and exhibits a certain degree of periodicity.

[0093] The pooling layer in the convolutional neural network is used to further reduce the dimension of the data and help retain the most discriminative feature information. During the pooling process, the CNN removes unnecessary details and redundant information by downsampling the features of the local area, while improving the network's robustness to deformation, noise, or changes. The feature vector set extracted by the convolutional neural network is used for subsequent equipment status judgment and fault prediction. In this technical solution, the convolutional neural network can effectively extract important features related to equipment failures through layer-by-layer convolution and pooling operations, thereby providing accurate data support for subsequent anomaly detection and fault prediction.

[0094] In a specific embodiment, before inputting the multidimensional feature vector set into the random forest model in step S3, the following steps are specifically included:

[0095] S31: Continuously sampling each type of feature data in the multidimensional feature vector set based on a sliding time window;

[0096] S32: Calculate the mean and standard deviation of each type of feature data within a preset window to construct a dynamic statistical feature set;

[0097] S33: When the deviation of the newly collected feature data from the mean value in the corresponding window exceeds a predetermined abnormal trigger threshold, the feature data is determined to be an abnormal warning point;

[0098] S34: The abnormal warning result is used as the prior judgment of the random forest model as input.

[0099] Specifically, to address the static nature of abnormal threshold settings and their poor adaptability to multiple operating conditions in traditional equipment status identification, a dynamic threshold adaptive judgment mechanism is introduced for the feature vector in step S3 before it is input into the random forest model. This mechanism continuously samples various monitoring feature data through a sliding time window, constructing dynamic mean and standard deviation indicators based on the historical statistical behavior of the same feature within the window interval to form a set of dynamic statistical features with time sensitivity. This feature set is continuously updated during actual operation, maintaining real-time characterization of equipment operating behavior at each point in time, enabling the system to adapt to the state changes of different equipment types, different load environments, and different operating time periods.

[0100] On this basis, the newly collected feature data is compared with the mean of that feature in the corresponding sliding window for deviation, and the deviation coefficient constructed based on the standard deviation is used to determine whether it exceeds the anomaly trigger range. If the deviation between the current data point and the statistical mean in the window exceeds the preset dynamic anomaly coefficient threshold, the data point is marked as an anomaly warning point, and an anomaly label vector is constructed. This label information is injected into the subsequent classification model as a priori features, and together with the original physical features, timestamp information, and geographic location features, it forms a high-dimensional input vector.

[0101] The above input vectors serve as input features for the random forest model. During model training and prediction, this dynamic anomaly prediction label, as one of the explicit classification dimensions, significantly improves the model's sensitivity to high-risk points, preventing blind or misjudgment of non-stationary fluctuations. Furthermore, a spatiotemporal clustering algorithm is incorporated into the model output backend to aggregate warning points with the same spatial or temporal distribution characteristics. This allows for the transition from identifying single-point anomalies to identifying regional or trending faults, improving the system's efficiency in identifying subtle, latent anomalies in complex operating environments.

[0102] This technical feature is functionally linked with the dynamic perception subsystem, feature fusion network, and fault prediction module, logically providing dynamic prior auxiliary judgment based on statistical behavior for model decision-making, forming a substantial improvement from the static threshold mechanism to the adaptive recognition mechanism. In the field of engineering equipment leasing, especially in scenarios with mixed deployment of multiple models and heterogeneous operation in multiple geographical regions, the use of this structure can effectively identify abnormal behavioral changes caused by differences in terrain, working conditions, or task loads, and output risk signals in a timely manner to provide data support for equipment maintenance and scheduling. As a manifestation of the deep integration of algorithms and monitoring technologies, this feature substantially improves the level of intelligent equipment status perception.

[0103] In a specific embodiment, determining whether the rental equipment is in an abnormal state in step S3 specifically includes the following steps:

[0104] S41: inputting the multidimensional feature vector set into a trained random forest model, where the trained random forest model is trained by a training dataset including historical operation data and corresponding status labels of the rental equipment at each usage stage;

[0105] S42: The random forest model performs preliminary classification processing based on the input feature data and outputs the predicted probability of each state category;

[0106] S43: Marking the data at the time point where the abnormality occurs based on the predicted probability;

[0107] S44: Based on the spatiotemporal clustering algorithm, perform clustering analysis on the data points marked as abnormal in the spatial and temporal dimensions to identify whether there are group, regional or stage-specific abnormal behaviors and obtain clustering results;

[0108] S45: Based on the clustering results, anomaly patterns of spatially continuous regions or time periods are formed;

[0109] S46: Compare the abnormal pattern with the common behavior pattern of the rental equipment in the historical data, calculate the similarity, and if the similarity is greater than a preset first threshold, mark it as an abnormal state.

[0110] Specifically, to accurately identify abnormal behavior in equipment operating status monitoring, an enhanced input vector containing time tags and geographic coordinates is constructed based on the multidimensional feature vector set output by a convolutional neural network. The time and location features are then integrated with the original operating condition features to serve as the joint discriminant input for the random forest model. The random forest model uses multiple decision tree sub-models to divide feature boundaries in different dimensions and uses ensemble learning to determine whether the equipment is in an abnormal state.

[0111] To further enhance the model's ability to aggregate and identify abnormal behavior across multiple devices and scenarios, a spatiotemporal clustering algorithm is introduced based on the classification results to process abnormal label sequences. A density-based clustering method is used to perform spatial aggregation analysis on data points marked as abnormal. This clustering of abnormal data with spatiotemporal concentration is grouped into the same abnormal event set, eliminating isolated points and misjudgments of outliers, and avoiding false alarms caused by occasional fluctuations. This processing mechanism incorporates an asynchronous processing module at the model output backend to automatically aggregate multiple abnormal judgment points within a time window and spatial neighborhood, forming a pattern of clustered abnormalities in spatially contiguous regions or time periods, which are then marked as high-risk events.

[0112] In actual applications, if a device experiences a sudden temperature rise and pressure fluctuation during continuous operation in a certain geographical area for a period of time, while similar situations do not occur in other areas, the spatiotemporal clustering mechanism can accurately classify such behaviors as abnormal events in a specific area, effectively avoiding misidentification due to equipment differences or diversity of deployment scenarios. This mechanism, together with the front-end feature extraction module, random forest classifier, and subsequent fault prediction logic, technically constitutes a functional association closed loop, significantly improving the system's ability to identify and respond to local or staged abnormal states in complex working conditions. As the core abnormality identification component in the overall monitoring system, the above-mentioned technical means not only enhances the spatial generalization performance of the classification model, but also realizes the visual expression of regional equipment health risks on the management platform through compatibility with structured geographic information systems, forming the basis for the intelligent perception and management linkage of the system of the present invention.

[0113] In a specific embodiment, the baseline model records the health parameter intervals of the rental equipment at different operating stages, which are used to identify trend changes in the time series.

[0114] Specifically, a baseline model is a mathematical model built based on the long-term historical operating data of a device. It aims to record and depict the various operating parameters of the device under normal operating conditions. These operating parameters include, but are not limited to, key indicators such as vibration, temperature, and pressure. The baseline model defines a healthy range or healthy interval for each operating phase of the device (such as the startup phase, stable operation phase, overload phase, etc.). This healthy interval is usually derived based on the device's historical data statistics and reflects the expected range of variation of these key parameters when the device is operating under normal conditions. By comparing it with the device's real-time operating data, the baseline model can help the system determine whether the device's status has deviated, thereby identifying potential failure risks. For example, when the device's vibration frequency or temperature exceeds the healthy interval defined by the baseline model, the system can issue an early warning, indicating that the device may be in an abnormal state or is about to fail.

[0115] In a specific embodiment, the fault prediction of the leased equipment in step S4 specifically includes the following steps:

[0116] S61: Use the anomaly detection results as explicit state variables in the input features of the LSTM model to identify short-term operational anomalies of the equipment.

[0117] S62: Slice the historical operating data of the rental equipment using fixed time windows to construct a multi-dimensional time series input matrix, where each row in the matrix corresponds to a time segment and each column corresponds to a different perception feature or state label;

[0118] S63: Input the time series input matrix into the LSTM model, and the LSTM model extracts the dynamic change characteristics of the long-term operation of the rental equipment;

[0119] S64: The comprehensive feature vector formed by combining the anomaly detection results, the dynamic change characteristics of historical operation data, and the baseline model is input into the LSTM model. Through its gating mechanism and memory function, the LSTM automatically captures abnormal trends and potential changes in the long-term operation of the equipment, thereby generating a hidden state vector.

[0120] S65: Based on the output of the LSTM model, the dynamic change characteristics of the leased equipment are compared with the healthy parameter range in the baseline model to calculate the failure risk of the equipment during the forecast period.

[0121] Specifically, anomaly detection results serve as one of the input features of the LSTM model, identifying short-term operational anomalies in the equipment and reflecting its health status over a short period of time, such as excessive temperature or vibration. These results serve as explicit state variables in the LSTM model, providing an immediate and direct fault marker for the model's time series prediction.

[0122] Historical operating data (such as equipment vibration, temperature, and pressure) is sliced ​​into fixed time windows to construct a multi-dimensional time series input matrix. The equipment status data within each time window forms the rows of the matrix. Each column corresponds to a different sensory feature or status label. Together, all this information forms a multi-dimensional input matrix. By feeding this constructed time series input matrix into the LSTM model, it begins modeling the equipment's long-term operating status, capturing the long-term dependencies of the time series data and automatically identifying dynamic characteristics of the equipment during long-term operation, such as gradually increasing temperature or vibration. The LSTM model not only receives historical data but also combines anomaly detection results and feature vectors generated by the baseline model with the equipment's dynamic characteristics to form a comprehensive feature vector, which is then input into the LSTM. Through its gating mechanism and memory function, the LSTM automatically captures the equipment's long-term operating trends and potential changes, generating a latent state vector that represents the "intrinsic representation" of the equipment's current state. When the equipment's dynamic characteristics deviate from the healthy parameter range defined by the baseline model, the LSTM model calculates a failure probability value, indicating the risk of equipment failure. This failure probability value is calculated by the LSTM output layer and combined with the baseline model to calibrate the risk value, further improving prediction accuracy. The output failure probability value is normalized to the range [0, 1] using a sigmoid mapping function, representing the likelihood of device failure. Based on this probability, the system can generate an alert and provide an estimate of the time and severity of the failure.

[0123] like Figure 4The figure below shows a performance comparison of fault prediction models. It compares the performance of different fault prediction models, including the baseline model, the LSTM model, and the combined baseline and LSTM model. The baseline model (dashed line) demonstrates the fault prediction capabilities of traditional methods. Prediction results fluctuate significantly, and in some cases, the model fails to accurately capture the moment of fault occurrence. The LSTM model (solid line) demonstrates the effectiveness of the LSTM model for fault prediction. The LSTM model provides smoother and more accurate predictions, particularly in detecting the fault location. The combined baseline and LSTM model (thick solid line) combines the strengths of the LSTM and baseline models, providing more accurate fault prediction results. The combined model generally more accurately identifies the moment of fault occurrence and reduces errors. Therefore, the combined model outperforms the traditional baseline model and the single LSTM model in terms of accuracy and robustness. It is particularly effective in identifying the moment of fault occurrence, reducing false positives and improving prediction accuracy. This provides reliable technical support for the operation and maintenance management of rental equipment, reducing equipment failures and improving the accuracy and efficiency of O&M.

[0124] In a specific embodiment, the electronic fence boundary is composed of polygonal geometric constraints dynamically generated by a cloud server. The dynamically generated polygonal geometric constraints are semantic fences constructed based on high-precision map data and can be adjusted in real time according to the actual operation area.

[0125] Specifically, to achieve dynamic adaptation of boundary control and trajectory deviation detection within the device's operating space, the cloud server uses a polygon boundary generation algorithm to construct a set of real-time electronic fence boundaries based on the device's current task parameters, historical trajectory data, and area configuration instructions. This boundary is composed of polygonal geometric constraints formed by sequentially combining a number of two-dimensional latitude and longitude coordinate points. The algorithm extracts the density of activity areas, boundary coverage, and the distribution of frequently visited areas from the device's historical operation trajectory, and combines this with the passable area standards defined by the operation type to form a set of dynamic boundary constraints that match the actual operating environment. Each boundary polygon in this set has a unique identifier and is bound to the corresponding device task.

[0126] When constructing the fence boundary, the cloud parses the task data issued by the device management platform, including the task number, land parcel, operation time period, and operation mode. It then injects parameters corresponding to the operation restrictions of the specified geographic area into the fence construction logic. The generated polygonal data structure contains not only the coordinates of the boundary points, but also attribute fields such as area type, restriction level, buffer radius, and other semantic information. The server transmits this geometric constraint data to the device in real time through the fence configuration interface. It also establishes a mapping between the task boundary and the device identifier in the database, and supports periodic refresh of the boundary and condition-triggered reconstruction.

[0127] During the movement of the equipment, the current positioning information collected can be projected in real time onto the two-dimensional plane defined by the above-mentioned fence boundary, and the spatial inclusion relationship between the current coordinates and the boundary polygon is calculated to determine whether it is within the legal operating area. At the same time, the trajectory point sequence output by the prediction module is also projected and checked point by point to see if it violates the spatial restrictions formed by the fence boundary. If the trajectory point falls outside the boundary, the trajectory deviation warning logic is triggered. The dynamic construction capability of the above-mentioned electronic fence can adapt to the geographical differences and task change requirements of different equipment in different operating scenarios. By combining cloud-based parameter parsing capabilities, historical trajectory modeling capabilities and task semantic configuration capabilities, the generation of fence boundaries not only has geometric accuracy, but also reflects the semantics of the operating context, thereby significantly enhancing the spatial compliance judgment capability of equipment operation behavior.

[0128] In a specific embodiment, the semantic fence constructed based on high-precision map data specifically includes the following steps:

[0129] S81: Constructing a rough boundary region based on the location data and historical operation data of the rental equipment;

[0130] S82: In the rough boundary area, based on the elevation data and image data of the working area where the rental equipment is located, extract the multidimensional spatial constraint factors in the area, the multidimensional spatial constraint factors including elevation gradient, slope, and obstruction information;

[0131] S83: Perform pixel-level classification and identification of multidimensional spatial constraint factors to identify inaccessible areas for rental equipment and obstacle information, and mark them as restricted areas to serve as spatial boundary constraints for semantic fences.

[0132] S84: Analyze the boundary information of the obstacle data and generate a restriction polygon that allows the rental equipment to move based on the historical operation behavior of the rental equipment;

[0133] S85: Integrate multiple restricted areas into the boundary of the restricted polygon to form a nested structure and obtain a semantic fence.

[0134] Specifically, in order to improve the spatial constraint accuracy of device trajectory deviation judgment and enhance the semantic expression ability of early warning, this application expands the traditional electronic fence mechanism based on static geometric boundaries to a semantic fence mechanism constructed based on high-precision map data.

[0135] The construction process begins by sourcing digital elevation model (DEM) and orthophoto (DOM) data containing terrain information for the geographic spatial extent of the rental equipment's operating area. Image recognition and geospatial analysis algorithms are then combined to identify physically inaccessible obstacle elements within the area, including buildings, mountains, walls, and traffic medians. This spatially constructs a multi-class obstacle boundary dataset. This obstacle boundary data is vectorized to generate a series of closed obstacle surface structures, each with a unique number and type. These surfaces are then merged with the polygons surrounding the equipment's historical operating trajectory to form a composite boundary that restricts the equipment's future path accessibility.

[0136] This composite boundary is defined as the main structure of the semantic fence, where the permitted operating area is represented as a primary polygon, and the obstacle surface is represented as a nested structure in the form of embedded polygons, forming a constraint boundary system that can express both spatial constraints and semantic attributes. The system supports task-level dynamic configuration and real-time refreshing of semantic fences. When the equipment's operating task is adjusted or the environment changes, the fence boundary can be automatically updated based on the latest map data and task templates, ensuring that the fence structure maintains dynamic consistency with the actual operating environment.

[0137] After the device state prediction module calculates future trajectory points, the system analyzes the spatial topology of the predicted trajectory points against the aforementioned semantic fence structure, including point-surface containment and point-boundary distance calculations. If any trajectory point falls into the impassable area defined by the obstacle surface or deviates significantly from the primary boundary, the system identifies a trajectory anomaly. These anomaly determinations are written into the early warning data structure, along with the coordinates of the offset point and the triggering obstacle surface identifier, enabling semantic-level description and traceability of trajectory deviation events.

[0138] This semantic fence technology feature, through fusion processing with map data, expands static spatial restrictions into a dynamic area model that includes terrain and building semantic labels. It effectively cooperates with device status data during the trajectory prediction result processing, which not only improves the practicality and accuracy of offset identification, but also enhances the semantic interpretation ability of early warning logic, thus forming an organic synergy between algorithm features and spatial modeling technology. It plays a core role in the overall monitoring system and has reusability, engineering adaptability and significant technological innovation.

[0139] In a specific embodiment, after step S5, the method further includes:

[0140] S6. Based on the trajectory deviation data in the early warning notification, the Euclidean distance is used to calculate the degree of deviation between the trajectory point and the electronic fence boundary. The trajectory deviation distribution and fault probability distribution are displayed on the visual interface through heat map rendering technology, realizing visual monitoring of the entire life cycle of the rental equipment.

[0141] S7. Generate a trajectory deviation distribution map and a failure probability distribution map of the rental equipment through heat map rendering to assist the management platform in visual analysis.

[0142] Specifically, to improve the interpretability and management response efficiency of trajectory deviations and equipment status anomalies, a visualization monitoring module is further set after step S5 to build a comprehensive display mechanism with Euclidean distance analysis and heat map rendering as the core. After the cloud server identifies the offset trajectory point based on the predicted trajectory and semantic fence structure, the system calculates the closest distance between the trajectory point and the semantic fence boundary line, uses Euclidean distance to measure the degree of trajectory deviation, and uses this value as a weight factor for the spatial deviation intensity to construct a heat map of the trajectory deviation distribution. At the same time, combined with the equipment failure probability value output by the LSTM model, the probability level corresponding to each time step also participates in the generation of heat map data as a weight dimension, forming a failure probability distribution map in the two-dimensional spatiotemporal domain.

[0143] The above two types of heat maps are embedded in the interactive interface of the management platform using standard visualization layer rendering. The trajectory deviation map expresses the deviation intensity through different color gradients, and the fault probability map expresses the evolution of fault trends through the density of node color blocks. Both can be superimposed or layered according to user selection, and support timeline dynamic playback function to achieve synchronous dynamic monitoring of equipment status and path deviation. Through this technical means, the management platform can quickly identify the severity and temporal characteristics of equipment deviation areas, assist decision makers in identifying potential operational risks and formulating intervention strategies, thereby forming a terminal visualization output mechanism that is functionally tightly coupled with the aforementioned prediction algorithm module, enhancing the system's overall real-time perception and response speed to fault risks and spatial deviations, and has significant practical value and system optimization effects.

[0144] The above describes a method for monitoring the entire life cycle of rental equipment based on the Internet of Things in an embodiment of the present application. The following describes a system for monitoring the entire life cycle of rental equipment based on the Internet of Things in an embodiment of the present application. Figure 5 In one embodiment of the present application, a full life cycle monitoring system for rental equipment based on the Internet of Things includes:

[0145] The acquisition module is used to monitor the operation process of the rental equipment through multi-source sensors and obtain real-time status data, including position data, vibration data, pressure data and temperature data;

[0146] The extraction module is used to extract features from real-time status data, obtain multiple feature vectors of rental equipment, and construct a multi-dimensional feature vector set;

[0147] The abnormality judgment module is used to input the multi-dimensional feature vector set into the random forest model and combine it with the spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state;

[0148] The fault diagnosis module is used to determine if the fault is true, and then combine the abnormal detection results, historical operation data and the preset baseline model to predict the fault of the rental equipment and obtain the fault probability value of the rental equipment;

[0149] The location prediction module is used to predict the future location of the rental equipment based on the location data and set the electronic fence boundary for the rental equipment. If the future location exceeds the electronic fence boundary, an early warning notification of trajectory deviation is generated and the warning information is pushed to the management platform.

[0150] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0151] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0152] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring the entire life cycle of rental equipment based on the Internet of Things, characterized in that: The method comprises: S1. Monitor the operation process of the rental equipment through multi-source sensors to obtain real-time status data, including position data, vibration data, pressure data, and temperature data; S2. Extract features from the real-time status data to obtain multiple feature vectors of the rental equipment and construct a multi-dimensional feature vector set; S3. Inputting the multidimensional feature vector set into a random forest model and combining it with a spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state; S4. If yes, then combine the abnormality detection results, historical operation data, and the preset baseline model to predict the failure of the leased equipment and obtain the failure probability value of the leased equipment; S5. Predict the future location of the rental device based on the location data, set an electronic fence boundary for the rental device, and generate a trajectory deviation warning notification if the future location exceeds the electronic fence boundary, and push the warning information to the management platform.

2. The method according to claim 1, characterized in that Extracting features from the real-time status data includes: Feature extraction is performed on the real-time status data based on a convolutional neural network model. The convolutional neural network model includes a deep feature fusion layer composed of multiple convolution and pooling structures, which is used to extract the frequency pattern of the vibration data, the mutation pattern of the pressure data, and the periodic pattern of the temperature data.

3. The method according to claim 1, characterized in that Before inputting the multidimensional feature vector set into the random forest model, the following steps are also included: S31: Continuously sampling each type of feature data in the multidimensional feature vector set based on a sliding time window; S32: Calculate the mean and standard deviation of each type of feature data within a preset window to construct a dynamic statistical feature set; S33: When the deviation of the newly collected feature data from the mean value in the corresponding window exceeds a predetermined abnormal trigger threshold, the feature data is determined to be an abnormal warning point; S34: The abnormal warning result is used as the prior judgment of the random forest model as input.

4. The method according to claim 1, wherein Determining whether the rental equipment is in an abnormal state includes: S41: inputting the multidimensional feature vector set into a trained random forest model, wherein the trained random forest model is trained by a training data set including historical operation data and corresponding status labels of the rental equipment at each usage stage; S42: The random forest model performs preliminary classification processing based on the input feature data and outputs the predicted probability of each state category; S43: Marking the data at the time point where the abnormality occurs based on the predicted probability; S44: Based on the spatiotemporal clustering algorithm, perform clustering analysis on the data points marked as abnormal in the spatial and temporal dimensions to identify whether there are group, regional or stage-specific abnormal behaviors and obtain clustering results; S45: Based on the clustering results, forming an abnormal pattern of a spatially continuous area or time period; S46: Compare the abnormal pattern with the common behavior pattern of the rental equipment in the historical data, calculate the similarity, and mark it as an abnormal state if the similarity is greater than a preset first threshold.

5. The method according to claim 1, wherein The baseline model records the health parameter intervals of the rental equipment at different operating stages, and is used to identify trend changes in the time series.

6. The method according to claim 1, characterized in that Failure prediction for rental equipment includes: S61: Use the anomaly detection results as explicit state variables in the input features of the LSTM model to identify short-term operational anomalies of the equipment. S62: Slice the historical operating data of the rental equipment using fixed time windows to construct a multi-dimensional time series input matrix, where each row in the matrix corresponds to a time segment and each column corresponds to a different perception feature or state label; S63: Inputting the time series input matrix into an LSTM model, wherein the LSTM model extracts dynamic change characteristics of the rental equipment during long-term operation; S64: The comprehensive feature vector formed by combining the anomaly detection results, the dynamic change characteristics of historical operation data, and the baseline model is input into the LSTM model. Through its gating mechanism and memory function, the LSTM automatically captures abnormal trends and potential changes in the long-term operation of the equipment, thereby generating a hidden state vector. S65: Based on the output of the LSTM model, the dynamic change characteristics of the leased equipment are compared with the healthy parameter range in the baseline model to calculate the failure risk of the equipment during the forecast period.

7. The method according to claim 1, characterized in that The electronic fence boundary is composed of polygonal geometric constraints dynamically generated by a cloud server. The dynamically generated polygonal geometric constraints are semantic fences constructed based on high-precision map data and can be adjusted in real time according to the actual operation area.

8. The method according to claim 7, characterized in that Semantic fences built based on high-precision map data include: S81: Constructing a rough boundary region based on the location data and historical operation data of the rental equipment; S82: Extracting, within the rough boundary area, multidimensional spatial constraint factors within the area based on elevation data and image data of the working area where the rental equipment is located, the multidimensional spatial constraint factors including elevation gradient, slope, and obstruction information; S83: performing pixel-level classification and identification on the multidimensional spatial constraint factors, identifying inaccessible areas for rental equipment and obstacle information, and marking them as restricted areas as spatial boundary constraint conditions of the semantic fence; S84: Analyze the boundary information of the obstacle data and generate a restriction polygon that allows the rental equipment to move based on the historical operation behavior of the rental equipment; S85: The semantic fence supports a nested structure, and multiple restricted areas are integrated into the boundary of the restricted polygon to form the nested structure, thereby obtaining the semantic fence.

9. The method according to claim 1, characterized in that After S5, it further includes: S6. Based on the trajectory deviation data in the warning notification, the degree of deviation between the trajectory point and the electronic fence boundary is calculated using Euclidean distance, and the trajectory deviation distribution and fault probability distribution are displayed on a visual interface using heat map rendering technology to achieve visual monitoring of the entire life cycle of the rental equipment; S7. Generate a trajectory deviation distribution map and a failure probability distribution map of the rental equipment through heat map rendering to assist the management platform in visual analysis.

10. A rental equipment full life cycle monitoring system based on the Internet of Things, used to implement a rental equipment full life cycle monitoring method based on the Internet of Things as described in any one of claims 1 to 9, characterized in that: The system comprises: An acquisition module is used to monitor the operation process of the rental equipment through multi-source sensors and obtain real-time status data, which includes position data, vibration data, pressure data, and temperature data; An extraction module, configured to extract features from the real-time status data, obtain multiple feature vectors of the rental equipment, and construct a multi-dimensional feature vector set; An abnormality judgment module, used to input the multidimensional feature vector set into a random forest model and combine it with a spatiotemporal clustering algorithm to determine whether the rental equipment is in an abnormal state; The fault diagnosis module is used to determine if the fault is true, and then combine the abnormal detection results, historical operation data and the preset baseline model to predict the fault of the rental equipment and obtain the fault probability value of the rental equipment; A location prediction module is used to predict the future location of the rental equipment based on the location data, set an electronic fence boundary for the rental equipment, and generate a trajectory deviation warning notification if the future location exceeds the electronic fence boundary, and push the warning information to the management platform.

Images