Electromechanical equipment management method and system based on Internet of Things

Abstract

The invention discloses an electromechanical equipment management method and system based on the Internet of Things, and belongs to the technical field of electromechanical equipment management, and the method comprises the steps: carrying out the preprocessing of original data, obtaining a first state vector, inputting the first state vector and a working condition label, outputting a risk score and a feature vector, and carrying out the calculation of a working condition label based on the risk score and the feature vector. Generating a first maintenance action, constructing a strategy function, executing the first maintenance action, collecting a current state vector of the equipment as a second state vector after execution is completed, constructing a feedback scoring function, inputting the second state vector to obtain a feedback score, and forming a training sample by the feedback score, the first maintenance action and the second state vector; and packaging the feedback score and the training sample into a block chain data unit, and realizing shared access through a key strategy. According to the invention, by combining edge calculation and an intelligent decision algorithm, real-time monitoring, fault prediction and intelligent maintenance decision are realized in equipment management.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromechanical equipment management, and in particular relates to an electromechanical equipment management method and system based on the Internet of Things. Background Art

[0002] With the continuous development of the Internet of Things (IoT), electromechanical equipment management has entered a new era. IoT technology provides real-time monitoring and management capabilities for equipment. Through the integration of various sensors and smart devices, it enables remote monitoring, data collection, and analysis. However, existing IoT device management systems still face challenges and shortcomings, particularly in areas such as equipment failure prediction, lifecycle management, and multi-device coordination. Most existing systems rely on traditional data collection and centralized processing, transmitting large amounts of data generated by devices to the cloud for analysis. However, this centralized processing approach suffers from latency issues. Especially when a large number of devices require real-time monitoring, network bandwidth and computing resource bottlenecks make it difficult for device management systems to ensure efficient responses. Furthermore, the diversity of devices and the complexity of their operating environments place higher demands on fault prediction and maintenance decision-making. Many existing systems utilize rule-based maintenance strategies that fail to dynamically optimize based on the real-time operating status of the equipment, resulting in inefficient equipment management, high maintenance costs, and prolonged equipment downtime. Furthermore, most existing technologies lack effective fault warning mechanisms and adaptive decision-making models, making them difficult to adapt to the operational needs of diverse equipment types and unable to provide effective warnings and scheduling before equipment failures occur.

[0003] Furthermore, existing systems generally suffer from data credibility and transparency issues. Traditional equipment management systems often rely on centralized data storage and management, which results in low device data security and susceptibility to tampering or loss. Ensuring the transparency and credibility of equipment maintenance records, fault histories, and repair plans is particularly challenging in multi-party environments. Because the equipment management process involves multiple parties, such as equipment manufacturers, operators, and repairers, existing systems often lack effective cross-departmental data sharing and transparency management. Unreliable data and a lack of transparency make it difficult for equipment management decisions to gain the trust and timely implementation of all parties involved, significantly impacting the efficiency and effectiveness of equipment management.

[0004] To this end, we propose an electromechanical equipment management method and system based on the Internet of Things to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems of data delay and low efficiency in the prior art, and to propose a method and system for managing electromechanical equipment based on the Internet of Things.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for managing electromechanical equipment based on the Internet of Things, comprising:

[0008] Collecting raw data, the raw data including a vibration signal, a current signal, a device housing temperature, and an operating condition label; preprocessing the raw data to obtain a first state vector;

[0009] Input a first state vector and a working condition label, and output a risk score and a feature vector. The risk score is obtained by remapping the historical state through an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state. The current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state.

[0010] Based on the risk score and the eigenvector, a first maintenance action is generated and a policy function is constructed. The first maintenance action is obtained through distribution in the action space, which is mapped into five levels of risk response actions. The policy function outputs a probability distribution in the action space through a two-branch network and introduces a policy smoothing term and an entropy penalty term.

[0011] Execute a first maintenance action, and after the execution is completed, collect the current state vector of the device as a second state vector, construct a feedback scoring function and input the second state vector into it to obtain a feedback score, and form a training sample with the feedback score, the first maintenance action, and the second state vector;

[0012] Feedback scores and training samples are encapsulated as blockchain data units, and shared access is achieved through key policies.

[0013] Preferably, preprocessing includes data windowing, data alignment and feature extraction.

[0014] Preferably, the first state vector includes the average value of the vibration signal, the standard deviation of the vibration signal, the average value of the current signal, the current change rate within the window time, the temperature average value and the temperature change rate.

[0015] Preferably, the fusion state realizes historical state remapping through an affine transformation mapping function based on the current working condition label. Each type of working condition corresponds to a set of linear parameters. The similarity between the current state and the historical state is calculated through the working condition-related matrix, and the weight is obtained through normalization processing, and finally the weighted sum of different historical states is obtained.

[0016] Preferably, the five-level risk response actions include: continuing operation; operating with a slight load reduction; scheduling manual inspection; automatically entering maintenance preparation state; and immediate shutdown.

[0017] Preferably, the dual-branch network includes a state branch and a risk branch, the state branch extracts action preferences through two linear layers plus ReLU, and the risk branch dynamically recalibrates the action preference weight through a Sigmoid modulation layer.

[0018] Preferably, the policy smoothing term represents the KL divergence between the current policy distribution and the distribution at the previous moment, and is used to keep the policy stable under stable working conditions.

[0019] Preferably, a feedback confidence weighting factor is introduced into the training sample, and the feedback confidence weighting factor is calculated based on the risk score and used for sampling during subsequent training.

[0020] An electromechanical equipment management system based on the Internet of Things, comprising:

[0021] A data acquisition module, the data acquisition module is used to collect multiple physical signals of the device and preprocess the multiple physical signals into a first state vector;

[0022] a risk assessment module, configured to generate a risk score and a feature vector using the first state vector and the operating condition label; the risk score is obtained by remapping the historical state using an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state; the current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state;

[0023] A strategy generation module, the strategy generation module is used to generate a first maintenance action and construct a strategy function to obtain a maintenance strategy distribution at a current moment;

[0024] an execution feedback module, wherein the execution feedback module executes the first maintenance action and collects a second state vector after the execution, constructs a feedback score and inputs the second state vector to obtain a feedback score, and forms a training sample with the feedback score, the first maintenance action, and the second state vector;

[0025] An information encapsulation and sharing module encapsulates feedback scores and training samples into blockchain data units and implements shared access through key policies.

[0026] In summary, the technical effects and advantages of the present invention are as follows: the electromechanical equipment management method and system based on the Internet of Things can significantly reduce the response delay in equipment management by performing data preprocessing and preliminary analysis on the edge computing node, ensuring that the equipment can be monitored and fault warnings can be performed in real time during actual operation. The introduction of edge computing not only solves the network bandwidth problem in the existing system, but also can dynamically optimize according to the real-time data and environmental changes of the equipment to ensure the efficiency and real-time performance of the equipment management system. In terms of equipment failure prediction, the present invention adopts an intelligent decision-making algorithm to provide more accurate fault warnings and maintenance recommendations by integrating the historical data, real-time data, equipment type, operating environment and other factors of the equipment. It also ensures that the data storage and sharing in the equipment management system have high transparency and credibility by introducing decentralized blockchain technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a structural schematic diagram of the present invention;

[0028] Figure 2 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0030] like Figure 1 As shown, a method for managing electromechanical equipment based on the Internet of Things includes:

[0031] Collecting raw data, the raw data including a vibration signal, a current signal, a device housing temperature, and an operating condition label; preprocessing the raw data to obtain a first state vector;

[0032] Input a first state vector and a working condition label, and output a risk score and a feature vector. The risk score is obtained by remapping the historical state through an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state. The current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state.

[0033] Based on the risk score and the eigenvector, a first maintenance action is generated and a policy function is constructed. The first maintenance action is obtained through distribution in the action space, which is mapped into five levels of risk response actions. The policy function outputs a probability distribution in the action space through a two-branch network and introduces a policy smoothing term and an entropy penalty term.

[0034] Execute a first maintenance action, and after the execution is completed, collect the current state vector of the device as a second state vector, construct a feedback scoring function and input the second state vector into it to obtain a feedback score, and form a training sample with the feedback score, the first maintenance action, and the second state vector;

[0035] Feedback scores and training samples are encapsulated as blockchain data units, and shared access is achieved through key policies.

[0036] The specific steps are as follows:

[0037] Step 1: Data collection and state representation construction

[0038] The goal of this step is to construct a unified, low-dimensional device state representation vector S that is suitable for downstream models by transforming the various physical signals collected by electromechanical equipment during operation into a unified, low-dimensional device state representation vector S t This is the input root node of the entire system, determining the accuracy and stability of downstream risk identification and maintenance strategy generation. Because the operating status of electromechanical equipment is affected by multiple factors (mechanical status, power characteristics, environmental factors, operating mode switching, etc.), traditional static statistical processing of multi-channel data suffers from problems such as insufficient expressiveness, time series distortion, and operating condition mismatch.

[0039] Therefore, this step innovatively proposes a fusion:

[0040] Asynchronous sampling signal alignment mechanism;

[0041] Sparse activation regularization structure based on dynamic selection of working conditions;

[0042] Disturbance suppression item (used to reduce false abnormal responses of equipment when switching between working conditions);

[0043] This will build a more stable, generalized equipment status representation that is adaptable to multiple operating environments.

[0044] Assume that the observation window at time t is the first 30 seconds and the sampling step is 10 seconds. The input raw data set is:

[0045] The three-axis vibration signal (1kHz, ADXL355 accelerometer) is synthesized by Euclidean method to

[0046] Current signal (1Hz, Hall transformer);

[0047] Equipment case temperature (0.2 Hz, thermocouple);

[0048] C t : Operating condition label (obtained from the PLC interface, identifying the current speed gear and load gear).

[0049] All data is collected locally by edge nodes (such as Jetson Nano) and stored as a sliding window D [t-30,t] , as the basis for state construction.

[0050] Step details

[0051] (1) Signal preprocessing and statistics within the window

[0052] Align and extract features for all input signals on the window [t-30,t]:

[0053] Acceleration signal a(t), calculate envelope mean μ a,t , standard deviation σ a,t ;

[0054] Current signal I(t), extract mean μ I,t Compared with the last 10s increment ΔI t ;

[0055] Temperature signal T(t), extract 30s mean μ T,t and the current instantaneous slope ΔT t .

[0056] Construct the initial feature representation vector:

[0057]

[0058] in:

[0059] μ a,t : The average value of the vibration signal in the window;

[0060] σ a,t : standard deviation of vibration signal;

[0061] μ I,t : average value of current signal;

[0062] Current change rate in the past 10 seconds;

[0063] μ T,t : average temperature;

[0064] Rate of temperature change.

[0065] (2) State mapping function design

[0066] In order to improve the adaptability of state representation to the actual behavior of the equipment, a working condition sensitivity structure mapping mechanism is introduced. The final state representation construction function is defined as follows:

[0067]

[0068] in:

[0069] The final state vector to be solved;

[0070] preliminary state feature vector;

[0071] λ1: sparse regularization coefficient (e.g. 0.01);

[0072] λ2: disturbance suppression regularization coefficient (e.g. 0.001);

[0073] β i (C t ): For the current working condition C t The sparse strength of the i-th state component (which can be defined by table lookup);

[0074] A set of specific dimension indexes, representing state dimensions that are sensitive to changes in operating conditions;

[0075] Represents the response gradient of state dimension j to the change of working condition label.

[0076] This optimization problem is implemented on edge devices through a lightweight optimizer with real-time processing capabilities.

[0077] Output

[0078] S t : The final state representation vector (dimension is 6), which provides input for the subsequent risk assessment module;

[0079] C t : The current equipment operating condition label, used for subsequent fusion model-assisted judgment and strategy personalized input.

[0080] Step 2: Risk assessment and dynamic feature fusion

[0081] This step, "Risk Assessment and Dynamic Feature Fusion", connects the previous and the next in the entire system. Its function is to build the state vector S based on step 1. t and Condition Label C t , output a real-time, continuous risk score R t , and generate the feature vector F that integrates the context and working condition factors t , which is used by the subsequent maintenance strategy generation module. Considering that the state distribution of electromechanical equipment under various operating conditions is highly heterogeneous, directly using the current state for classification will lead to frequent false alarms. Therefore, we designed a lightweight model structure that combines the temporal attention mechanism, the working condition adaptive affine transformation and the rare abnormality constraint term, so that the risk score is not only based on the current state S t, and can also integrate contextual historical status and form stronger recognition capabilities for abnormal samples in special working conditions, thereby systematically improving the stability and practical usability of risk scoring.

[0082] enter

[0083] This step uses all the output from step 1:

[0084] Structured state vector, including vibration, current, temperature and other characteristics;

[0085] The equipment's current operating status tag is uploaded by the PLC system at a fixed frequency;

[0086] The state cache of the past k=3 time steps is stored in the edge device memory.

[0087] First, to improve the model's ability to utilize historical information without increasing computational costs, we designed a weighted fusion mechanism based on condition gating. This mechanism dynamically activates historical state information by modeling the structural correlation between the current state and historical states:

[0088]

[0089] in:

[0090] It is the historical state after integration;

[0091] Based on the current working condition C t Affine transformation mapping function, each type of working condition corresponds to a set of linear parameters

[0092] is the attention matching matrix related to the working condition;

[0093] γ i is the activation weight for each history segment;

[0094] All parameters are obtained through model training, and the preloaded model structure is called based on the working conditions.

[0095] The innovation of this structure lies in Realize working condition-aware historical state remapping, so that historical states can participate in scoring and judgment from different "perspectives" under different working conditions, and improve the system's ability to express complex dynamic working conditions.

[0096] Next, concatenate the current state with the fusion state to form an intermediate representation Input to the risk scoring submodule. To address the problem of weak representation of high-risk rare events in the sample distribution, we designed a sparse activation regularization term in the scoring function to prevent the model from overfitting to high-frequency low-risk samples:

[0097]

[0098] in:

[0099] is the fully connected network weight;

[0100] σ(·) is the Sigmoid function, output R t ∈(0,1) represents the risk score of the device;

[0101] The last term is the rare anomaly activation constraint term. When the overall state amplitude is small, this term amplifies the output, which is conducive to the identification of weak anomalies.

[0102] λ is the regularization coefficient, ∈ prevents division by zero, and is typically set to 1e -6 .

[0103] This structure makes the system more sensitive in judging high-risk, low-frequency fault events, and not be "masked" by stable operation under most working conditions.

[0104] All models run on the edge, and each round of calculation is completed within 80ms, fully meeting industrial deployment requirements.

[0105] Output

[0106] R t : Risk score result, which is a continuous value and indicates the probability of the device currently failing (for direct use by the subsequent policy judgment module);

[0107] F t :The intermediate features of the state are fused and input into the strategy optimization model of the next step, providing a contextual expression basis for strategy generation.

[0108] This step innovatively embeds the working condition structure information deeply into the weighted fusion and affine projection of historical states, and combines it with a sparse activation mechanism to achieve risk score output with high robustness, high sensitivity, and high generalization ability in scenarios with limited edge resources.

[0109] Step 3: Reinforcement learning strategy generation

[0110] This step, “Reinforcement Learning Strategy Generation”, is a key step in the patent system to complete the transition from “condition perception and risk assessment” to “active decision-making and maintenance action”. The goal is to generate a risk score R based on the output of the previous step. t and the eigenvector F t , automatically generate the maintenance strategy action A of the current devicet , and construct the strategy function π(F t ,R t ) for subsequent learning and iterative optimization. In actual IoT device management scenarios, the operating status and external environment of devices are highly volatile, maintenance resources are often limited, and false triggering or delayed maintenance can be costly. Therefore, this step proposes a reinforced policy network that combines a risk segmentation policy response mechanism with policy confidence entropy constraints. This achieves adaptive hierarchical response of maintenance policies and policy stability regulation, ensuring a dynamic balance between real-time performance, energy efficiency, and operational effectiveness.

[0111] First, considering the risk score R t The interval characteristics (ie: R t The larger the value, the more dangerous it is), we will take the action space Mapped into five levels of risk response actions, namely:

[0112] a1: continue running;

[0113] a2: Slightly reduced load operation;

[0114] a3: dispatch manual inspection;

[0115] a4: Automatically enter the maintenance preparation state;

[0116] a5: Shut down immediately.

[0117] We designed a policy function π(F t ,R t ), which is used to fuse the vector F according to the current state t and risk score R t , output the probability distribution in the action space, and sample through the distribution to obtain the current strategy action A t . This function is implemented through a two-branch network:

[0118] Status branch: Input F t , action preference is extracted through two linear layers plus ReLU;

[0119] Risk branch: Input R t , after a Sigmoid modulation layer, the action preference weight is dynamically recalibrated for risk;

[0120] The final output action distribution π(F t ,R t ) is normalized by Softmax.

[0121] The confidence entropy suppression term is used in the policy optimization phase to prevent the model from having overly dispersed policy outputs (high uncertainty) or overly aggressive behavior in low-risk states. The policy loss we designed is:

[0122]

[0123] in:

[0124] The first term is the expected maintenance cost loss, where C(A t ) represents the cost constant corresponding to the selected action, for example, C(a1) = 0, C(a5) = 10;

[0125] The second term is the policy smoothing term, where KL(·) represents the KL divergence between the current policy distribution and the distribution at the previous moment, which encourages the policy to remain stable when the working conditions are stable.

[0126] The third term is the entropy penalty term, only in R t It is activated when it is less than the risk threshold θ, which is used to suppress the policy uncertainty in the low-risk state and reduce the policy fuzzy area ( represents the entropy of the policy distribution);

[0127] λ1,λ2 are regularization coefficients, θ is the risk tolerance threshold (e.g., θ = 0.35);

[0128] The core innovation of the above loss design lies in: it not only considers the direct loss of the current action, but also embeds the stability of strategy changes and operational controllability in low-risk scenarios, strengthening the system's ability to rationally select strategies in "non-emergency situations".

[0129] This structure uses the proximal policy optimization algorithm (PPO) for iterative optimization during the training phase and the distilled version during the deployment phase. The model inference speed is controlled within 50ms and is adapted to the edge node environment.

[0130] Output

[0131] The optimal maintenance action selected at the current moment is given by the policy distribution π(F t ,R t ) is sampled;

[0132] π(F t ,R t ): The maintenance strategy distribution at the current moment, used for the next step execution and feedback training to participate in the update.

[0133] Step 4: Execute control and feedback learning

[0134] This step "Execution Control and Feedback Learning" is responsible for strengthening the maintenance action A output by the previous step strategy module. tIt is actually executed in the device system and the subsequent device operation status is collected to build a feedback score r for strategy training optimization. t .

[0135] First execute control action A t , the actual method is as follows:

[0136] For action a1 (continue operation) or a2 (load reduction operation), the edge node sends a power or speed setting command to the PLC via the Modbus-TCP or OPC UA interface;

[0137] For a3 (manual notification), the maintenance scheduling module is called to generate a maintenance work order and send a text message or App notification;

[0138] For a4 (maintenance preparation) or a5 (shutdown), an execution instruction is sent to the equipment safety power-off module through the serial port, and a fan / oil pump shutdown command is added if necessary.

[0139] After the action is executed, the system re-collects the current state vector S from the device after a fixed delay Δt (such as 60 seconds) t+Δt The acquisition method is the same as step 1 (the edge node extracts the sensor value through MODBUS polling and assembles it into S t+Δt vector).

[0140] We construct a feedback scoring function r t , evaluate the execution effect:

[0141] r t =(||S t ||2-||S t+Δt ||2)-β·C(A t )

[0142] in:

[0143] ||S t ||2 and ||S t+Δt ||2 are the state modulus before and after the action, representing the overall fluctuation amplitude of the device. The smaller the value, the more stable it is.

[0144] C(A t ) is the estimated maintenance cost of the current action (e.g., C(a1) = 0, C(a5) = 10);

[0145] β is the cost adjustment coefficient, which is empirically set to 0.3-0.5.

[0146] Then, the score is combined with the current action and state features to form a training sample:

[0147] τ t =(F t ,Rt ,A t ,r t )

[0148] The sample has a complete structure and clear temporal logic, which is the strategy network π(F t ,R t ) provides a feedback basis for the training input.

[0149] In order to make the system respond faster to high-risk situations and prevent sample redundancy, we introduce the feedback confidence weighting factor w t , used for sampling during subsequent training:

[0150] w t =γ·R t +(1-γ)

[0151] in:

[0152] R t is the risk score;

[0153] γ∈[0,1], it is recommended to be set to 0.7 to give high-risk samples priority in training;

[0154] The implementation method is that the samples in the sample buffer are w t Participate in strategy training batch construction as sampling weights.

[0155] Finally, all samples τ t The samples are stored in a circular cache of the edge device (e.g., a FIFO buffer with a length of 1000). Whenever the number of samples meets the policy update condition (e.g., 32), a policy update training will be triggered.

[0156] Output

[0157] r t : The feedback score of this round of maintenance action, which indicates the net benefit of this execution behavior on the system status;

[0158] τ t : Standard feedback sample unit for periodic training of reinforcement strategy network.

[0159] Step 5: Blockchain records are shared with multiple parties

[0160] This step encapsulates the maintenance feedback and policy execution sample information generated in the previous step into a trusted structured form and writes it to the blockchain network. This ensures an immutable record of device maintenance activities, auditable processes, and traceable events. Furthermore, considering the multiple parties involved in the device management system, including operators, manufacturers, and third-party maintenance providers, we have designed a sharing mechanism based on access policies to ensure authorized data access by different roles, supporting multi-party collaboration throughout the device lifecycle.

[0161] First, τ t Encapsulate and construct data structures for on-chain Its fields are as follows:

[0162] action_id:A t , action number;

[0163] risk_score:R t , risk scoring before strategy execution;

[0164] reward:r t , feedback rating;

[0165] policy_id: by π(F t ,R t ) The version hash in this round is obtained and the strategy version is marked;

[0166] state_fingerprint: The following compression network ψ(·) is used to calculate F t and R t After the joint encoding, SHA256 hash is performed to obtain the status summary.

[0167] The compression network ψ(·) is configured as follows:

[0168] The input dimension is 13 (12-dimensional F t +1 Dimension R t );

[0169] The first layer is a linear transformation, with an 8-dimensional output and a ReLU activation function;

[0170] The second layer is a linear transformation that outputs 4 dimensions as a summary of the state representation;

[0171] Finally, perform a standard hash: SHA256(ψ(F t ,R t ))Form state_fingerprint.

[0172] Then construct the feedback value weight ω of the data t , used for on-chain priority judgment:

[0173]

[0174] in:

[0175] The first one reflects the effect of the strategy;

[0176] The second item reflects the on-chain priority of high-risk executions;

[0177] λ is the high-risk enhancement factor, with a recommended value of 1.0;

[0178] θ is the high-risk judgment threshold, and 0.6 is recommended.

[0179] The system uses a lightweight consensus mechanism (such as PBFT) for sorted writing. Each edge node submits all data to be uploaded to the chain every ΔT seconds. By sorting ω t Arrange in descending order, the first K items are put into the block, and the rest are queued or discarded.

[0180] Finally, to achieve multi-party controllable sharing, we Attach an access permission tag κ t , is composed as follows:

[0181]

[0182] Indicates that the data is encrypted and marked by the node using the public key of partner j;

[0183] Written into the smart contract control field access_policy, reading is allowed only when the authorized party's address matches.

[0184] This step will maintain the behavior result r t and the complete feedback sample τ t Encapsulated as blockchain data units Through the feedback scoring driven sorting consensus mechanism and compressed state summary technology, the on-chain writing efficiency is effectively improved, and multi-party controllable access is achieved through key strategies. It is the basic guarantee for achieving transparency, trusted sharing, and accurate traceability of device life cycle data.

[0185] The technical solutions in the above-mentioned embodiments of the present application have at least the following technical effects or advantages: This solution can significantly reduce the response delay in equipment management by performing data preprocessing and preliminary analysis on edge computing nodes, ensuring that the equipment can be monitored and fault warnings can be performed in real time during actual operation. The introduction of edge computing not only solves the network bandwidth problem in existing systems, but also can dynamically optimize according to the real-time data and environmental changes of the equipment to ensure the efficiency and real-time performance of the equipment management system. In terms of equipment fault prediction, the present invention adopts an intelligent decision-making algorithm to provide more accurate fault warnings and maintenance recommendations by integrating historical data, real-time data, equipment type, operating environment and other factors of the equipment. It also ensures that the data storage and sharing in the equipment management system are highly transparent and credible by introducing decentralized blockchain technology.

[0186] The present application also provides a method and system for managing electromechanical equipment based on the Internet of Things. Figure 2 As shown, an electromechanical equipment management system based on the Internet of Things includes:

[0187] A data acquisition module, the data acquisition module is used to collect multiple physical signals of the device and preprocess the multiple physical signals into a first state vector;

[0188] a risk assessment module, configured to generate a risk score and a feature vector using the first state vector and the operating condition label; the risk score is obtained by remapping the historical state using an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state; the current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state;

[0189] A strategy generation module, the strategy generation module is used to generate a first maintenance action and construct a strategy function to obtain a maintenance strategy distribution at a current moment;

[0190] an execution feedback module, wherein the execution feedback module executes the first maintenance action and collects a second state vector after the execution, constructs a feedback score and inputs the second state vector to obtain a feedback score, and forms a training sample with the feedback score, the first maintenance action, and the second state vector;

[0191] An information encapsulation and sharing module encapsulates feedback scores and training samples into blockchain data units, drives sorting consensus and status summary through feedback scores, and realizes shared access through key policies.

[0192] The working principle is as follows: by performing data preprocessing and preliminary analysis on edge computing nodes, the response delay in device management is reduced; edge computing is introduced and dynamic optimization is performed based on the real-time data of the device and environmental changes; by integrating the device's historical data, real-time data, as well as factors such as device type and operating environment, more accurate fault warnings and maintenance recommendations are provided; and decentralized blockchain technology is introduced to ensure information storage and sharing.

[0193] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for managing electromechanical equipment based on the Internet of Things, characterized in that: include: Collecting raw data, including vibration signals, current signals, device housing temperature, and operating condition labels; Preprocessing the original data to obtain a first state vector; Input a first state vector and a working condition label, and output a risk score and a feature vector. The risk score is obtained by remapping the historical state through an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state. The current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state. Based on the risk score and the eigenvector, a first maintenance action is generated and a policy function is constructed. The first maintenance action is obtained through distribution in the action space, which is mapped into five levels of risk response actions. The policy function outputs a probability distribution in the action space through a two-branch network and introduces a policy smoothing term and an entropy penalty term. Execute a first maintenance action, and after the execution is completed, collect the current state vector of the device as a second state vector, construct a feedback scoring function and input the second state vector into it to obtain a feedback score, and form a training sample with the feedback score, the first maintenance action, and the second state vector; Feedback scores and training samples are encapsulated as blockchain data units, and shared access is achieved through key policies.

2. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The preprocessing includes data windowing, data alignment and feature extraction.

3. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The first state vector includes the average value of the vibration signal, the standard deviation of the vibration signal, the average value of the current signal, the current change rate within the window time, the temperature average value and the temperature change rate.

4. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The fusion state realizes historical state remapping through an affine transformation mapping function based on the current working condition label. Each type of working condition corresponds to a set of linear parameters. The similarity between the current state and the historical state is calculated through the working condition-related matrix, and the weight is obtained through normalization processing. Finally, the weighted sum of different historical states is obtained.

5. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The five-level risk response actions include: continuing operation; operating with a slight load reduction; scheduling manual inspection; automatically entering maintenance preparation status; and immediate shutdown.

6. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The dual-branch network includes a state branch and a risk branch. The state branch extracts action preferences through two linear layers plus ReLU, and the risk branch dynamically recalibrates the action preference weight through a Sigmoid modulation layer.

7. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: The policy smoothing term represents the KL divergence between the current policy distribution and the distribution at the previous moment, and is used to keep the policy stable under stable working conditions.

8. The electromechanical equipment management method based on the Internet of Things according to claim 1, characterized in that: A feedback confidence weighting factor is introduced into the training sample, and the feedback confidence weighting factor is calculated based on the risk score and is used for sampling during subsequent training.

9. An electromechanical equipment management system based on the Internet of Things, characterized in that: include: A data acquisition module, the data acquisition module is used to collect multiple physical signals of the device and preprocess the multiple physical signals into a first state vector; a risk assessment module, configured to generate a risk score and a feature vector using the first state vector and the operating condition label; the risk score is obtained by remapping the historical state using an affine transformation mapping function, and the feature vector is obtained by concatenating the current state and the fused state; the current state is obtained from the original data, and the fused state is obtained by dynamically activating the historical state; A strategy generation module, the strategy generation module is used to generate a first maintenance action and construct a strategy function to obtain a maintenance strategy distribution at a current moment; an execution feedback module, wherein the execution feedback module executes the first maintenance action and collects a second state vector after the execution, constructs a feedback score and inputs the second state vector to obtain a feedback score, and forms a training sample with the feedback score, the first maintenance action, and the second state vector; An information encapsulation and sharing module encapsulates feedback scores and training samples into blockchain data units and implements shared access through key policies.

Images