Mechanical metering equipment intelligent management method and system based on Internet of Things

Abstract

The invention discloses a mechanical metering equipment intelligent management method and system based on the Internet of Things. Multi-source metering equipment data are acquired through a sensor group; performing data preprocessing on the multi-source metering equipment data to obtain fused metering equipment data; constructing a physical constraint LSTM network based on a coulomb friction model, embedding a kinetic equation as a regularization item, and extracting multi-source feature time sequence association by using a Transform encoder to obtain an equipment state prediction model; outputting the health index of the metering equipment; and generating metering equipment archive information according to the metering equipment health index, and predicting an equipment calibration period and generating an equipment maintenance strategy according to the metering equipment archive information. Calibration period prediction and maintenance strategy optimization are realized, and the excessive maintenance cost is reduced.

Description

Technical Field

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

[0002] In modern industrial production and scientific research, mechanical metrology equipment serves as a core tool to ensure the accuracy of measurement of physical quantities such as force, torque, and pressure. Its operating status directly affects product quality, the reliability of experimental results, and production safety. However, traditional mechanical metrology equipment management methods face many challenges. Traditional calibration and maintenance strategies often rely on fixed cycles or manual inspections, and there are problems with excessive or untimely maintenance in the process of equipment information management and equipment file establishment. Frequent calibration of equipment in good health increases operating costs, and the failure to intervene in potentially faulty equipment in a timely manner may lead to major downtime accidents or the risk of measurement inaccuracy. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an intelligent management method and system for mechanical metrology equipment based on the Internet of Things.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is as follows: further, in the above-mentioned intelligent management method of mechanical metrology equipment based on the Internet of Things, the intelligent management method of mechanical metrology equipment includes the following steps: The sensor group acquires vibration data, current data, and temperature data of the mechanical metrology equipment body, and simultaneously collects equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data; performing data preprocessing on the multi-source metrology equipment data to obtain standard metrology equipment data, and fusing feature tensors in the standard metrology equipment data using a feature fusion model weighted by an Attention mechanism to obtain fused metrology equipment data; Based on the Coulomb friction model, a physical constraint LSTM network is constructed, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the temporal correlation of multi-source features to obtain the LSTM-Transformer equipment state prediction model. Inputting the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification, and outputting a metering equipment health index; Measuring equipment archive information is generated according to the measuring equipment health index, and equipment calibration cycle is predicted and equipment maintenance strategy is generated according to the measuring equipment archive information.

[0005] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, the data preprocessing of the multi-source metrology equipment data to obtain standard metrology equipment data includes: Calculating the mean and standard deviation of the multi-source metering equipment data, identifying data points that deviate from the mean by more than 3 times the standard deviation as outliers, and repairing the outliers using adjacent effective value interpolation to obtain repaired equipment data; Using a Butterworth low-pass filter to filter the high-frequency signal data of vibration and current in the repair equipment data; using a moving average filter method to smooth the low-frequency slowly varying data of temperature, humidity and electromagnetic interference to obtain smoothed equipment data; The smoothed device data is uniformly converted using a minimum-maximum normalization method to obtain standard measurement device data.

[0006] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, the feature fusion model weighted by the Attention mechanism is used to fuse the feature tensors in the standard metrology equipment data to obtain fused metrology equipment data, including: Convert the standard metrology data into feature tensors and input the feature tensors into the Attention mechanism; each attention head maps the input feature tensors into query, key, and value respectively through linear transformation; Use the dot product attention mechanism to calculate the similarity between the query and the key to obtain the attention score matrix; Perform weighted summation of the values ​​according to the attention score matrix to obtain the output of each attention head; The output of each attention head is concatenated and then converted into a fused feature tensor through a linear transformation layer to obtain the fused metering equipment data.

[0007] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, a physical constraint LSTM network is constructed based on the Coulomb friction model, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the temporal association of multi-source features to obtain the LSTM-Transformer equipment state prediction model, including: The LSTM network includes at least an input gate, a forget gate, an output gate, and a cell state. On the LSTM network, the dynamic equation corresponding to the Coulomb friction model is added to the loss function as a regularization term.

[0008] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, the physical constraint LSTM network is constructed based on the Coulomb friction model, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the time series correlation of multi-source features to obtain the LSTM-Transformer equipment state prediction model, which also includes: The Transformer encoder structure consists of multiple stacked encoding layers, each of which contains a multi-head self-attention mechanism and a feedforward neural network, and uses residual connections and layer normalization techniques; The input is time series data that has undergone feature fusion. The multi-head self-attention mechanism is used to capture the long-distance temporal correlation between different time steps. The feedforward neural network performs nonlinear transformation on the attention output to extract high-order features.

[0009] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, the inputting of the fused metrology equipment data into the LSTM-Transformer equipment status prediction model for identification and outputting the metrology equipment health index includes: The model is trained end-to-end, concatenating the output of the LSTM network and the output of the Transformer encoder, and outputting the device state prediction result through a fully connected layer. The loss function is the sum of the mean square error (MSE) loss and the regularization term of the dynamic equation. The fused metrology equipment data is input into the trained LSTM-Transformer equipment status prediction model. The model outputs the component status indicators of the equipment in the current state, and generates the metrology equipment health index based on the component status indicators.

[0010] Furthermore, in the above-mentioned intelligent management method for mechanical metrology equipment based on the Internet of Things, generating metrology equipment archive information according to the metrology equipment health index, predicting the equipment calibration cycle and generating the equipment maintenance strategy according to the metrology equipment archive information include: The archival information of measuring equipment shall at least include equipment information, operating data, health status index and calibration and maintenance records; The Cox proportional hazards model was used to analyze the impact of factors on the attenuation of equipment measurement accuracy and predict the calibration cycle of equipment under different health conditions, taking equipment health index, age, operation frequency and environmental factors as covariates.

[0011] Furthermore, in the intelligent management system for mechanical metrology equipment based on the Internet of Things, the intelligent management system for mechanical metrology equipment includes the following modules: The equipment data acquisition module is used to obtain vibration data, current data, and temperature data of the mechanical metrology equipment through a sensor group. It also collects equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data. A feature tensor fusion module is used to perform data preprocessing on the multi-source metrology equipment data to obtain standard metrology equipment data, and to fuse the feature tensors in the standard metrology equipment data using a feature fusion model weighted by an Attention mechanism to obtain fused metrology equipment data; The prediction model building module is used to build a physical constraint LSTM network based on the Coulomb friction model, embed the dynamic equation as a regularization term, and use the Transformer encoder to extract the temporal correlation of multi-source features to obtain the LSTM-Transformer device state prediction model; An equipment status prediction module, configured to input the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification and output a metering equipment health index; The equipment intelligent management module is used to generate metering equipment archive information according to the metering equipment health index, predict the equipment calibration cycle and generate equipment maintenance strategy according to the metering equipment archive information.

[0012] Furthermore, in the intelligent management system for mechanical metrology equipment based on the Internet of Things, the equipment data acquisition module includes the following submodules: A repair submodule is used to calculate the mean and standard deviation of the multi-source metering equipment data, identify data points that deviate from the mean by more than 3 times the standard deviation as outliers, and repair the outliers using the adjacent effective value interpolation method to obtain repaired equipment data; a smoothing submodule for filtering the high-frequency signal data of vibration and current in the repair equipment data using a Butterworth low-pass filter; and smoothing the low-frequency slowly varying data of temperature, humidity, and electromagnetic interference using a moving average filter to obtain smoothed equipment data; A submodule is obtained, which is used to uniformly convert the smoothed device data using a minimum-maximum normalization method to obtain standard measurement device data.

[0013] Furthermore, in the intelligent management system for mechanical metrology equipment based on the Internet of Things, the equipment data acquisition module also includes the following submodules: The conversion submodule is used to convert the standard metrology equipment data into feature tensors and input the feature tensors into the Attention mechanism; each attention head maps the input feature tensor into query, key and value respectively through linear transformation; The calculation submodule is used to calculate the similarity between the query and the key using the dot product attention mechanism to obtain the attention score matrix; A summation submodule, configured to perform weighted summation of the values ​​according to the attention score matrix to obtain the output of each attention head; The splicing submodule is used to splice the output of each attention head, and then convert the output into a fused feature tensor through a linear transformation layer to obtain the fused metering equipment data.

[0014] Its beneficial effects are: 1. It can adaptively mine the correlation between different data, and compared with traditional single data or simple weighted analysis, it can more accurately reflect the actual operating status of the equipment and improve the accuracy of fault diagnosis; 2. In the case of sudden changes in equipment operating conditions or complex fault scenarios, it can improve the physical consistency of the prediction results and effectively reduce the false alarm and missed alarm rate; 3. It breaks through the limitations of traditional methods in capturing long-distance time series dependencies. It can quickly identify the complex relationships between different time steps and different physical quantities during the operation of the equipment, and improve the ability of mechanical metrology equipment to capture fault signs. 4. It can achieve calibration cycle prediction and maintenance strategy optimization, reduce excessive maintenance costs, shorten equipment downtime, and significantly improve equipment availability and management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0016] Figure 1 Schematic diagram of a first embodiment of an intelligent management method for mechanical metrology equipment based on the Internet of Things in an embodiment of the present invention; Figure 2 Schematic diagram of a second embodiment of an intelligent management method for mechanical metrology equipment based on the Internet of Things in an embodiment of the present invention; Figure 3 This is a schematic diagram of the first embodiment of the intelligent management system for mechanical metrology equipment based on the Internet of Things in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0019] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the intelligent management method of mechanical metrology equipment based on the Internet of Things includes the following steps: Step 101: Acquire vibration data, current data, and temperature data of the mechanical metrology equipment body through a sensor group, and simultaneously collect equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data; Specifically, 1. Sensor group selection and deployment; Vibration data acquisition: A three-axis accelerometer (ADXL345, with a range of ±16g, 13-bit resolution, and a sampling frequency of up to 3200Hz) is selected and installed on key vibration transmission path nodes such as the bearing seat and gearbox of the mechanical metrology equipment. Magnetic or bolt fixing methods are used to ensure a rigid connection between the sensor and the equipment body, thereby accurately obtaining the equipment vibration characteristics in the time and frequency domains. Current data acquisition: A closed-loop Hall current sensor (CSM100A, measuring range 0-100A, accuracy 0.5%, bandwidth 20kHz) is used. It is connected in series to the three-phase power supply line of the equipment motor and connected through a dedicated current transformer terminal. Temperature Data Collection: Surface-mount thermocouple sensors (K-type thermocouples, measuring temperature range -40°C to 1200°C, accuracy of ±1.5°C) are attached to the surfaces of heat-prone components of the equipment, such as the motor windings, bearings, and power modules. Thermal grease is applied between the sensor and the component to improve temperature measurement accuracy. A temperature sensor (DS18B20, measuring temperature range -55°C to 125°C, accuracy of ±0.5°C) is installed in the equipment control cabinet to monitor the overall operating temperature of the equipment. Operation Frequency Data Collection: Micro switches or proximity sensors (Omron D2SW, lifespan of over 1 million operations) are installed inside or near the equipment's operating components, such as buttons, switches, and knobs. By detecting the action signals of these components, the frequency and type of equipment operations, such as calibration, power on / off, and parameter setting, are counted. Ambient temperature and humidity data collection: Use an integrated temperature and humidity sensor (SHT30, temperature measurement range -40°C to 125°C, accuracy of ±0.3°C, humidity measurement range 0% to 100%RH, accuracy of ±2%RH) and deploy it in the working environment of the device, 1-2 meters away from the device, at the same height as the device operation panel to avoid direct sunlight and heat dissipation, to collect ambient temperature and humidity data in real time. Electromagnetic interference data collection: An electromagnetic interference sensor (HMS-901, measuring frequency range 10 kHz-3 GHz, sensitivity no higher than -50 dBm) is used. It is installed within 1-3 meters of the device, with the sensor antenna perpendicular to the ground and facing the direction of possible electromagnetic interference sources. 2. Data acquisition hardware architecture; Build an IoT-based sensor network using a star topology. Each sensor connects to a data acquisition module (an STM32F767 development board with an ARM Cortex-M4 core) via bus interfaces such as RS-485, I2C, and SPI. The data acquisition module includes a built-in real-time operating system (RTOS, such as FreeRTOS) to synchronize and cache data from multiple sensors. The data acquisition module also transmits collected data to a cloud server or local data processing center via wireless communications such as Wi-Fi, 4G / LTE, or Ethernet wired communication.

[0020] Step 102: Preprocess the multi-source metrology equipment data to obtain standard metrology equipment data, and fuse the feature tensors in the standard metrology equipment data using a feature fusion model weighted by the Attention mechanism to obtain fused metrology equipment data. Specifically, in this embodiment, the mean and standard deviation of multi-source metering equipment data are calculated, and data points that deviate from the mean by more than 3 times the standard deviation are identified as outliers. The outliers are repaired using the adjacent effective value interpolation method to obtain repaired equipment data; The high-frequency signal data of vibration and current in the repair equipment data are filtered using a Butterworth low-pass filter; the low-frequency slowly varying data of temperature, humidity and electromagnetic interference are smoothed using a moving average filter to obtain smoothed equipment data; The minimum-maximum normalization method is used to uniformly transform the smoothed equipment data to obtain the standard measurement equipment data.

[0021] Convert the standard metering equipment data into feature tensors and input the feature tensors into the Attention mechanism; each attention head maps the input feature tensors into queries, keys, and values ​​through linear transformations; Use the dot product attention mechanism to calculate the similarity between the query and the key to obtain the attention score matrix; The output of each attention head is obtained by weighted summing of the values ​​according to the attention score matrix; The output of each attention head is concatenated and then converted into a fused feature tensor through a linear transformation layer to obtain the fused metering equipment data.

[0022] Specifically, 1. Data preprocessing process; Data cleaning: Outlier detection: For each sensor's collected time series data, the statistically based 3σ principle (Z-score method) is applied to first calculate the data's mean and standard deviation. Data points that deviate from the mean by more than three standard deviations are identified as outliers and corrected using interpolation of adjacent valid values. For operation frequency data, if the same operation is recorded repeatedly within a very short period of time (less than 50ms), it is deleted as an erroneous operation. Noise filtering: For high-frequency signal data such as vibration and current, a Butterworth low-pass filter (the cutoff frequency is determined by the device's natural frequency and operating frequency range, such as 1000 Hz) is used for filtering to remove high-frequency noise interference. For low-frequency, slowly varying data such as temperature, humidity, and electromagnetic interference, a moving average filter method (with a window size set to 5 sampling points) is used for smoothing. Data normalization: We use the minimum-maximum normalization method to convert the data collected by each sensor to a range of 0 to 1. This method subtracts the minimum value from each sensor's data and then divides it by the difference between the maximum and minimum values. For discrete count data such as operation frequency, we use logarithmic normalization, which takes the logarithm of the data and then normalizes it to prevent zero values ​​from affecting subsequent calculations. 2. Feature fusion model with weighted Attention mechanism; Model Architecture: The model consists of an input layer, multiple attention heads, a feature fusion layer, and an output layer. The input layer receives preprocessed standard metering equipment data and converts it into a feature tensor. This tensor contains information about the time series length and feature dimensions. The feature dimensions cover various characteristics such as vibration, current, temperature, operation frequency, ambient temperature and humidity, and electromagnetic interference. Attention mechanism implementation: Compute query, key, and value: For each attention head, the input feature tensor is mapped to query, key, and value respectively through three different linear transformations. Calculate attention score: Use the dot product attention mechanism to calculate the similarity between the query and the key to obtain the attention score matrix. Weighted summation: Based on the attention score matrix, the values ​​are weighted summed to obtain the output of each attention head. Feature fusion: The outputs of multiple attention heads are concatenated and then converted into a fused feature tensor through a linear transformation layer. This enables adaptive weighted fusion of different sensor features and highlights key features that have a greater impact on device status.

[0023] Step 103: Build a physical constraint LSTM network based on the Coulomb friction model, embed the dynamic equation as a regularization term, use the Transformer encoder to extract the multi-source feature time series correlation, and obtain the LSTM-Transformer device state prediction model; Specifically, in this embodiment, the LSTM network includes at least an input gate, a forget gate, an output gate, and a cell state. On the LSTM network, the dynamic equation corresponding to the Coulomb friction model is added to the loss function as a regularization term.

[0024] The Transformer encoder structure consists of multiple stacked encoding layers, each of which contains a multi-head self-attention mechanism and a feedforward neural network, and uses residual connections and layer normalization techniques; The input is time series data that has undergone feature fusion. The multi-head self-attention mechanism is used to capture the long-distance temporal correlation between different time steps. The feedforward neural network performs nonlinear transformation on the attention output to extract high-order features.

[0025] Specifically, 1. Physically constrained LSTM network architecture design; 1. Physical modeling of Coulomb friction model; For the moving parts (bearings, gear pairs) of mechanical metrology equipment, the Coulomb friction characteristics are abstracted into physical constraints: Friction force calculation: The magnitude of the friction force is proportional to the normal load on the contact surface, and its direction is opposite to the relative motion speed. When the component is stationary, the friction force is equal to the driving force (static friction state), and the friction coefficient remains constant during movement (kinetic friction state).

[0026] Fusion of dynamic equations: Introducing friction loss terms into the equipment dynamics model to describe the relationship between component displacement, velocity, acceleration, friction, and driving force, forming a state-space equation that includes friction effects as a physical constraint on equipment motion.

[0027] 2. Enhanced LSTM network structure; Input layer design: The input dimensions include the original sensor features (vibration, current, temperature, etc.) and the physical model-derived features (theoretical acceleration and friction power loss calculated according to the dynamic equations), which are spliced ​​together to form a multidimensional input vector.

[0028] Cell state expansion: Physical state variables (theoretical displacement and velocity of components) are added to the traditional LSTM cell state, so that the cell state contains both data-driven features and physical model parameters.

[0029] Regularization term embedding: Adding a physical constraint term to the loss function forces the model-predicted device state (vibration acceleration) to be consistent with the theoretical calculated value based on the Coulomb friction model. The specific implementation is as follows: Based on the current input characteristics, the Coulomb friction model is used to calculate theoretical acceleration, friction and other physical quantities; The square of the difference between the vibration acceleration predicted by the model and the theoretical acceleration is used as a regularization term, and the weight is adjusted through hyperparameters to balance the priority of data fitting and physical constraints.

[0030] 3. Key parameter configuration; Hidden layer dimensions: Set 2-3 LSTM layers based on device complexity, with 128-256 hidden units per layer, optimized through grid search; Time step: Match the sensor sampling frequency (100 Hz sampling corresponds to a time step of 0.01 seconds), and the window length is set to 3-5 times the typical vibration period of the equipment (1024 time steps for rotating machinery); Initialization of the initial state: Using the calibration data of the device when it is running without load, the initial displacement and velocity are calculated as the initial values ​​of the LSTM cell state.

[0031] 2. Transformer encoder temporal association modeling; 1. Multi-source feature temporal alignment; Timestamp synchronization: For sensor data with different frequencies, such as vibration and current, the sampling frequency is unified (100Hz) through interpolation to generate a multi-dimensional time series with equal time intervals; Position encoding: Sine-cosine position encoding is used to generate an encoding vector containing absolute position information for each time step. This vector is added to the feature tensor and then input into the Transformer to solve the sequential dependency problem of time series data.

[0032] 2. Optimization of multi-head self-attention mechanism; Cross-sensor attention: When calculating the attention score, different sensor features (frequency domain features of vibration signals and harmonic components of current signals) are allowed to be correlated with each other, capturing the coupling effects of multiple physical quantities in device operation; Local-global attention: Two attention heads are set: one focuses on the local dependencies of adjacent time steps (window size 16), and the other captures the long-range dependencies of the entire sequence. Weighted fusion improves the ability to capture temporal correlations. Masking mechanism: Add a mask to the time steps where no data is collected (device downtime period) to prevent invalid data from participating in attention calculation.

[0033] 3. Feedforward network and layer normalization; Dual-channel feedforward network: Design two parallel sub-networks to perform nonlinear transformations on the attention output: Channel 1: Enhance high-frequency features (impact components of vibration signals) using the ReLU activation function; Channel 2: retains low-frequency trends (temperature slow-changing features) and uses the Sigmoid activation function; The output is the element-by-element addition of the two-channel results, which improves the feature expression capability.

[0034] Layer normalization improvements: Adaptive normalization parameters are added after each encoding layer to dynamically adjust the normalization mean and variance based on the device type (pressure sensor, torque sensor), enhancing model generalization.

[0035] 3. LSTM-Transformer fusion strategy; 1. Heterogeneous feature fusion architecture; Early fusion: The LSTM output (including time series features and physical constraint information) and the Transformer output (multi-source time series correlation features) are concatenated before entering the fully connected layer to form a composite feature vector that includes time series dynamic characteristics, physical constraint characteristics, and cross-sensor correlation characteristics. Adaptive weight allocation: Through a learnable gating mechanism, the weights of LSTM and Transformer outputs are dynamically adjusted according to the device operating status (normal / warning / fault). For example, the weight of physical constraint features is increased in the fault state.

[0036] Step 104: Input the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification, and output the metering equipment health index; Specifically, in this embodiment, the model is trained end-to-end, the output of the LSTM network and the output of the Transformer encoder are spliced ​​together, and the device state prediction result is output through the fully connected layer. The loss function is the sum of the MSE mean square error loss and the regularization term of the dynamic equation; The fused metrology equipment data is input into the trained LSTM-Transformer equipment status prediction model. The model outputs the component status indicators of the equipment in the current state, and generates the metrology equipment health index based on the component status indicators.

[0037] Specifically, an end-to-end training approach is used, concatenating the output of the LSTM network and the output of the Transformer encoder, and outputting the device state prediction results through a fully connected layer. The loss function is derived by adding the mean squared error loss to the regularization term of the dynamics equation. The optimizer uses the Adam optimization algorithm, with an initial learning rate set to 0.001 and dynamically adjusted as the training progresses. Early stopping is used to prevent model overfitting, and training is terminated when the validation set loss no longer decreases. Equipment Health Index Identification: Feature-fused metering equipment data is fed into a trained LSTM-Transformer equipment status prediction model. The model then outputs status indicators for key components of the equipment in its current state, such as bearing wear, probability of abnormal motor winding temperature, and gearbox vibration severity. The weights of each indicator are determined using the Analytic Hierarchy Process (AHP) to construct an equipment health index calculation model. The equipment health index ranges from 0 to 100, with higher values ​​indicating better equipment health. The weights of each indicator are determined through expert scoring and statistical analysis. Each indicator is also normalized to produce a score on a scale of 0 to 100.

[0038] Step 105: Generate metering equipment archive information based on the metering equipment health index, predict the equipment calibration cycle and generate an equipment maintenance strategy based on the metering equipment archive information.

[0039] Specifically, in this embodiment, the metering equipment archive information includes at least equipment information, operating data, health status index, and calibration and maintenance records; The Cox proportional hazards model was used to analyze the impact of factors on the attenuation of equipment measurement accuracy and predict the calibration cycle of equipment under different health conditions, taking equipment health index, age, operation frequency and environmental factors as covariates.

[0040] Specifically, 1. Equipment file information generation; The device profile information includes the following: Basic information: record equipment model, number, manufacturer, date of manufacture, installation location, measurement standards, etc. Operation data: stores historical vibration data, current data, temperature data, operation frequency data, ambient temperature and humidity data, electromagnetic interference data, etc., saved in time series format, and provides data query and visualization functions. Health status: Record the historical curve of equipment health index, the evolution trend of key component status, and fault history (including fault time, fault type, repair measures, etc.). Calibration and maintenance records: record calibration time, calibration results, calibration personnel; maintenance time, maintenance content, maintenance personnel, replacement parts information, etc. 2. Calibration cycle prediction; A calibration cycle prediction model was established based on the equipment health index and historical calibration data. The Cox proportional hazards model in survival analysis was used, with the equipment health index, age, operation frequency, and environmental factors as covariates. The impact of these factors on the attenuation of equipment measurement accuracy was analyzed, and the calibration cycle of the equipment under different health conditions was predicted. When the equipment health index dropped to the warning threshold (70), a calibration reminder was triggered. Combined with the predicted calibration cycle and equipment usage frequency, a reasonable calibration plan was arranged. 3. Maintenance strategy generation; Generate personalized maintenance strategies based on equipment health index and fault diagnosis results: Preventive maintenance: For equipment with a health index of 80-100, develop a regular maintenance plan, such as cleaning and lubricating the equipment once a month, and checking the sensor connection status and equipment grounding once a quarter. Predictive maintenance: When the equipment health index drops to 70-80, the system prepares repair parts in advance and arranges targeted maintenance measures based on the model's predicted fault type and remaining service life. For example, it replaces bearings that are about to fail and repairs gears with minor wear. Fault repair: When the equipment health index is lower than 70 or a serious fault occurs, a fault alarm is immediately triggered and a detailed fault diagnosis report (including fault location, fault cause, impact, etc.) is generated to guide maintenance personnel to perform rapid repairs, record the fault handling process, and update equipment file information.

[0041] Its beneficial effects are: 1. It can adaptively mine the correlation between different data, and compared with traditional single data or simple weighted analysis, it can more accurately reflect the actual operating status of the equipment and improve the accuracy of fault diagnosis; 2. In the case of sudden changes in equipment operating conditions or complex fault scenarios, it can improve the physical consistency of the prediction results and effectively reduce the false alarm and missed alarm rate; 3. It breaks through the limitations of traditional methods in capturing long-distance time series dependencies. It can quickly identify the complex relationships between different time steps and different physical quantities during the operation of the equipment, and improve the ability of mechanical metrology equipment to capture fault signs. 4. It can achieve calibration cycle prediction and maintenance strategy optimization, reduce excessive maintenance costs, shorten equipment downtime, and significantly improve equipment availability and management efficiency.

[0042] See also Figure 2 In the intelligent management method of mechanical metrology equipment based on the Internet of Things, the feature fusion model with weighted attention mechanism is used to fuse the feature tensors in the standard metrology equipment data. The fused metrology equipment data is obtained by the following steps: Step 201: Convert the standard metering equipment data into a feature tensor and input the feature tensor into the Attention mechanism; each attention head maps the input feature tensor into query, key, and value respectively through linear transformation; Step 202: Calculate the similarity between the query and the key using the dot product attention mechanism to obtain an attention score matrix; Step 203: Perform weighted summation on the values ​​according to the attention score matrix to obtain the output of each attention head; Step 204: Concatenate the outputs of each attention head and convert them into fused feature tensors through a linear transformation layer to obtain fused metering equipment data.

[0043] The above is an introduction to the embodiment of the intelligent management method of mechanical metrology equipment based on the Internet of Things of the present invention. Figure 3 In the intelligent management system of mechanical metrology equipment based on the Internet of Things, the intelligent management system of mechanical metrology equipment includes the following modules: The equipment data acquisition module is used to obtain vibration data, current data, and temperature data of the mechanical metrology equipment through a sensor group. It also collects equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data. The feature tensor fusion module is used to preprocess multi-source metrology equipment data to obtain standard metrology equipment data. The feature fusion model weighted by the Attention mechanism is used to fuse the feature tensors in the standard metrology equipment data to obtain fused metrology equipment data. The prediction model building module is used to build a physical constraint LSTM network based on the Coulomb friction model, embed the dynamic equation as a regularization term, and use the Transformer encoder to extract the temporal correlation of multi-source features to obtain the LSTM-Transformer device state prediction model; The equipment status prediction module is used to input the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification and output the metering equipment health index; The equipment intelligent management module is used to generate metering equipment archive information based on the metering equipment health index, predict the equipment calibration cycle and generate equipment maintenance strategy based on the metering equipment archive information.

[0044] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent management method for mechanical metrology equipment based on the Internet of Things, characterized in that: The intelligent management method of mechanical metrology equipment comprises the following steps: The sensor group acquires vibration data, current data, and temperature data of the mechanical metrology equipment body, and simultaneously collects equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data; performing data preprocessing on the multi-source metrology equipment data to obtain standard metrology equipment data, and fusing feature tensors in the standard metrology equipment data using a feature fusion model weighted by an Attention mechanism to obtain fused metrology equipment data; Based on the Coulomb friction model, a physical constraint LSTM network is constructed, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the temporal correlation of multi-source features to obtain the LSTM-Transformer equipment state prediction model. Inputting the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification, and outputting a metering equipment health index; Measuring equipment archive information is generated according to the measuring equipment health index, and equipment calibration cycle is predicted and equipment maintenance strategy is generated according to the measuring equipment archive information.

2. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, characterized in that: The step of preprocessing the multi-source metering equipment data to obtain standard metering equipment data includes: Calculating the mean and standard deviation of the multi-source metering equipment data, identifying data points that deviate from the mean by more than 3 times the standard deviation as outliers, and repairing the outliers using adjacent effective value interpolation to obtain repaired equipment data; Using a Butterworth low-pass filter to filter the high-frequency signal data of vibration and current in the repair equipment data; using a moving average filter method to smooth the low-frequency slowly varying data of temperature, humidity and electromagnetic interference to obtain smoothed equipment data; The smoothed device data is uniformly converted using a minimum-maximum normalization method to obtain standard measurement device data.

3. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, characterized in that: The feature fusion model weighted by the Attention mechanism is used to fuse the feature tensors in the standard metrology equipment data to obtain fused metrology equipment data, including: Convert the standard metrology data into feature tensors and input the feature tensors into the Attention mechanism; each attention head maps the input feature tensors into query, key, and value respectively through linear transformation; Use the dot product attention mechanism to calculate the similarity between the query and the key to obtain the attention score matrix; Perform weighted summation of the values ​​according to the attention score matrix to obtain the output of each attention head; The output of each attention head is concatenated and then converted into a fused feature tensor through a linear transformation layer to obtain the fused metering equipment data.

4. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, wherein: The physical constraint LSTM network is constructed based on the Coulomb friction model, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the time series association of multi-source features to obtain the LSTM-Transformer device state prediction model, which includes: The LSTM network includes at least an input gate, a forget gate, an output gate, and a cell state. On the LSTM network, the dynamic equation corresponding to the Coulomb friction model is added to the loss function as a regularization term.

5. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, characterized in that: The physical constraint LSTM network is constructed based on the Coulomb friction model, the dynamic equation is embedded as a regularization term, and the Transformer encoder is used to extract the time series association of multi-source features to obtain the LSTM-Transformer device state prediction model, which also includes: The Transformer encoder structure consists of multiple stacked encoding layers, each of which contains a multi-head self-attention mechanism and a feedforward neural network, and uses residual connections and layer normalization techniques; The input is time series data that has undergone feature fusion. The multi-head self-attention mechanism is used to capture the long-distance temporal correlation between different time steps. The feedforward neural network performs nonlinear transformation on the attention output to extract high-order features.

6. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, characterized in that: The step of inputting the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification and outputting a metering equipment health index includes: The model is trained end-to-end, concatenating the output of the LSTM network and the output of the Transformer encoder, and outputting the device state prediction result through a fully connected layer. The loss function is the sum of the mean square error (MSE) loss and the regularization term of the dynamic equation. The fused metrology equipment data is input into the trained LSTM-Transformer equipment status prediction model. The model outputs the component status indicators of the equipment in the current state, and generates the metrology equipment health index based on the component status indicators.

7. The method for intelligent management of mechanical metrology equipment based on the Internet of Things according to claim 1, wherein: Generating the metering equipment archive information according to the metering equipment health index, predicting the equipment calibration cycle and generating the equipment maintenance strategy according to the metering equipment archive information, includes: The archival information of measuring equipment shall at least include equipment information, operating data, health status index and calibration and maintenance records; The Cox proportional hazards model was used to analyze the impact of factors on the attenuation of equipment measurement accuracy and predict the calibration cycle of equipment under different health conditions, taking equipment health index, age, operation frequency and environmental factors as covariates.

8. Intelligent management system for mechanical metrology equipment based on the Internet of Things, characterized by: The intelligent management system for mechanical metrology equipment includes the following modules: The equipment data acquisition module is used to obtain vibration data, current data, and temperature data of the mechanical metrology equipment through a sensor group. It also collects equipment operation frequency data, ambient temperature and humidity data, and electromagnetic interference data to obtain multi-source metrology equipment data. A feature tensor fusion module is used to perform data preprocessing on the multi-source metrology equipment data to obtain standard metrology equipment data, and to fuse the feature tensors in the standard metrology equipment data using a feature fusion model weighted by an Attention mechanism to obtain fused metrology equipment data; The prediction model building module is used to build a physical constraint LSTM network based on the Coulomb friction model, embed the dynamic equation as a regularization term, and use the Transformer encoder to extract the temporal correlation of multi-source features to obtain the LSTM-Transformer device state prediction model; An equipment status prediction module, configured to input the fused metering equipment data into the LSTM-Transformer equipment status prediction model for identification and output a metering equipment health index; The equipment intelligent management module is used to generate metering equipment archive information according to the metering equipment health index, predict the equipment calibration cycle and generate equipment maintenance strategy according to the metering equipment archive information.

9. The intelligent management system for mechanical metrology equipment based on the Internet of Things according to claim 8, characterized in that: The device data acquisition module includes the following submodules: A repair submodule is used to calculate the mean and standard deviation of the multi-source metering equipment data, identify data points that deviate from the mean by more than 3 times the standard deviation as outliers, and repair the outliers using the adjacent effective value interpolation method to obtain repaired equipment data; a smoothing submodule for filtering the high-frequency signal data of vibration and current in the repair equipment data using a Butterworth low-pass filter; and smoothing the low-frequency slowly varying data of temperature, humidity, and electromagnetic interference using a moving average filter to obtain smoothed equipment data; A submodule is obtained, which is used to uniformly convert the smoothed device data using a minimum-maximum normalization method to obtain standard measurement device data.

10. The intelligent management system for mechanical metrology equipment based on the Internet of Things according to claim 8, characterized in that: The device data acquisition module also includes the following submodules: The conversion submodule is used to convert the standard metrology equipment data into feature tensors and input the feature tensors into the Attention mechanism; each attention head maps the input feature tensor into query, key and value respectively through linear transformation; The calculation submodule is used to calculate the similarity between the query and the key using the dot product attention mechanism to obtain the attention score matrix; A summation submodule, configured to perform weighted summation of the values ​​according to the attention score matrix to obtain the output of each attention head; The splicing submodule is used to splice the output of each attention head, and then convert the output into a fused feature tensor through a linear transformation layer to obtain the fused metering equipment data.

Images