Internet of Things equipment prediction system

Abstract

The invention relates to the technical field of Internet of Things, and discloses an Internet of Things equipment prediction system. Comprising an equipment output prediction module for predicting a production output prediction result of target equipment by using a pre-trained output prediction model, a maintenance period prediction module for predicting a maintenance period prediction result of the target equipment by using a pre-trained maintenance period prediction model, and an equipment fault prediction module for predicting the production output of the target equipment. The fault information prediction module is used for predicting a fault information prediction result of target equipment by utilizing a pre-trained equipment fault prediction model, and the equipment spare part prediction module is used for predicting a spare part demand prediction result of the target equipment by utilizing a pre-trained equipment spare part prediction model; and utilizing a pre-trained equipment energy consumption prediction model to predict an energy consumption parameter prediction result of the target equipment. According to the invention, comprehensive prediction of the Internet of Things equipment can be realized through each prediction module.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and more specifically to an IoT device prediction system. Background Technology

[0002] With the booming development of smart manufacturing, the intelligent transformation of IoT devices has become a key factor for enterprises to enhance their competitiveness. IoT device systems are platforms used for centralized monitoring, management, and maintenance of production equipment, enabling remote control, status monitoring, data collection, and analysis. Currently, the detection functions of IoT device systems are relatively limited, failing to provide comprehensive testing of production equipment. Summary of the Invention

[0003] In view of this, the present invention provides an Internet of Things (IoT) device prediction system to solve the problem that the detection function of IoT device management systems is relatively simple and cannot perform comprehensive detection of production equipment.

[0004] According to a first aspect, the present invention provides an Internet of Things (IoT) device prediction system, comprising:

[0005] The equipment output prediction module is used to predict the production output of the target equipment based on the equipment detection parameters, equipment output parameters, daily operation and maintenance parameters and equipment fault parameters of the target equipment, using a pre-trained output prediction model.

[0006] The maintenance cycle prediction module is used to predict the maintenance cycle of the target equipment based on the current environmental parameters and equipment detection parameters of the target equipment, using a pre-trained maintenance cycle prediction model.

[0007] The equipment fault prediction module is used to predict the fault information of the target equipment based on the equipment detection parameters and daily operation and maintenance parameters of the target equipment, using a pre-trained equipment fault prediction model.

[0008] The equipment spare parts prediction module is used to predict the spare parts demand of the target equipment based on the equipment testing parameters, daily operation and maintenance parameters, production task parameters and current environmental parameters of the target equipment, using a pre-trained equipment spare parts prediction model.

[0009] The equipment spare parts prediction module is used to predict the energy consumption parameters of the target equipment based on the equipment detection parameters, current environmental parameters, and daily energy consumption parameters of the target equipment, using a pre-trained equipment energy consumption prediction model.

[0010] The IoT device prediction system in this invention can achieve comprehensive prediction of production equipment through the equipment output prediction module, maintenance cycle prediction module, equipment failure prediction module, equipment spare parts prediction module, and equipment spare parts prediction module.

[0011] In some optional embodiments, the equipment output prediction module, maintenance cycle prediction module, equipment failure prediction module, equipment spare parts prediction module, and equipment energy consumption prediction module of the present invention all include corresponding data processing submodules, feature extraction submodules, model training submodules, model optimization submodules, and model prediction submodules.

[0012] This invention improves the accuracy of prediction results through its various sub-modules, including equipment output prediction module, maintenance cycle prediction module, equipment failure prediction module, equipment spare parts prediction module, and equipment energy consumption prediction module.

[0013] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0014] The data acquisition module is used to collect equipment testing parameters, equipment output parameters, daily operation and maintenance parameters, and equipment fault parameters of the target equipment.

[0015] This invention can obtain various detection parameters of the Internet of Things device prediction system through the data acquisition module.

[0016] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0017] The equipment data display module is used to display the production output forecast, maintenance cycle forecast, fault information pre-handling results, spare parts demand forecast, and energy consumption parameter forecast of the target equipment.

[0018] This invention can clearly present various prediction results to users through the device data display module.

[0019] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0020] The equipment fault alarm module is used to issue fault alarms to the target equipment based on the predicted fault information.

[0021] This invention, through its equipment fault alarm module, facilitates the timely generation of alarm information.

[0022] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0023] The equipment safety execution module is used to execute safety measures when there are safety hazards in the target equipment.

[0024] This invention, through its device safety execution module, helps to ensure the safety of the target device.

[0025] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0026] The device interface management module is used to match the target interface and target interface protocol according to the device type of the target device.

[0027] This invention facilitates the successful establishment of communication connections by the target device through the device interface management module.

[0028] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0029] The user management module is used to manage user permission information and user personal information.

[0030] This invention facilitates the management of user information through a user management module.

[0031] In some optional embodiments, the IoT device prediction system of the present invention further includes:

[0032] The equipment resource management module is used to manage equipment resources.

[0033] This invention facilitates the management of equipment resources through an equipment resource management module.

[0034] In some optional embodiments, the IoT device prediction system of the present invention further includes: a device performance evaluation module, used to evaluate the overall performance of the target device based on the production output prediction results, maintenance cycle prediction results, fault information prediction results, spare parts demand prediction results, and energy consumption parameter prediction results of the target device.

[0035] This invention, through its equipment performance evaluation module, facilitates timely assessment of the overall performance of the target equipment. Attached Figure Description

[0036] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is a structural block diagram of an Internet of Things (IoT) device prediction system according to an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] This embodiment provides an IoT device prediction system, such as Figure 1 As shown, it includes: equipment output prediction module 101, maintenance cycle prediction module 102, equipment failure prediction module 103, equipment spare parts prediction module 104, and equipment energy consumption prediction module 105.

[0040] The equipment output prediction module mentioned above is used to predict the production output of the target equipment based on the equipment detection parameters, equipment output parameters, equipment operation and maintenance parameters, and equipment fault parameters of the target equipment, using a pre-trained output prediction model.

[0041] Specifically, the target equipment includes, but is not limited to, production equipment such as testing equipment, communication equipment, hydraulic press equipment, and heating equipment. Equipment testing parameters include, but are not limited to, spindle speed, feed rate, runtime, average processing speed, number of downtimes, pressure, flow rate, oil temperature, and vibration. Equipment output parameters include, but are not limited to, product order parameters and material supply parameters. Routine maintenance parameters for the target equipment include, but are not limited to, failure occurrence time t. f Fault type T f Faulty part P f Maintenance measures M f Replacement of parts C p and repair time D m Detailed information is required. In addition, the daily operation and maintenance parameters for the target equipment also include the equipment maintenance cycle T. c Maintenance Content C c Inspection Report R i wait.

[0042] In a specific example, the equipment output prediction module includes: a data processing submodule, a feature extraction submodule, a model training submodule, a model optimization submodule, and a model prediction submodule.

[0043] The data processing submodule within the equipment output prediction module processes equipment detection parameters, output parameters, maintenance parameters, and fault parameters of the target equipment. For example, data cleaning algorithms are used to process raw data at edge computing nodes or in the cloud. For equipment operation data within the equipment detection parameters, the 3σ criterion combined with sliding window technology is used to identify and correct outliers. For instance, if the spindle speed of a vertical machining center momentarily exceeds three times the normal average standard deviation and does not return to normal within a certain time window (e.g., 5 minutes), based on the equipment status log, if it is a sensor malfunction, it is marked as an outlier and corrected according to historical normal data from the same period; if it is a temporary overload due to processing special workpieces, it is retained. Simultaneously, sensors are automatically calibrated periodically, adjusting the sensor's measurement accuracy based on standard source signals to ensure data accuracy.

[0044] The feature extraction submodule within the equipment output prediction module extracts features closely related to output from various preprocessed data sets. Besides the direct operating parameters of the target equipment, it also includes derived features, such as the material removal rate calculated based on spindle speed and feed rate, and the tool wear rate estimated based on tool change frequency and machining time. Min-Max normalization or Z-score standardization methods are used to unify the dimensions of feature data at different scales. For example, spindle speed and machining time are normalized to the 0-1 range so that subsequent machine learning models can treat each feature equally and avoid model training bias due to differences in dimensions.

[0045] The model training submodule within the equipment output prediction module trains the model on features closely related to output after feature extraction. For example, based on the characteristics of equipment operation data and the needs of output prediction, a suitable artificial intelligence model is selected. For data with obvious time-series characteristics, such as long-term accumulated daily equipment output data, a Long Short-Term Memory (LSTM) network is chosen, which can effectively capture long-term dependencies in the data and cope with seasonal and periodic changes in the production process. If the data exhibits complex nonlinear relationships and has high feature dimensionality, a Multilayer Perceptron (MLP) neural network is selected. It constructs complex function mappings through multiple layers of neurons, which can mine the potential correlation between output and various factors from multi-source data.

[0046] The preprocessed data is divided into training and test sets according to a certain ratio. For example, 80% of the training set is selected for training, and 20% is selected for testing. The selected model is trained using the training set, and reasonable hyperparameters are set, such as the number of hidden layers and neurons in LSTM, and the number of layers and nodes per layer in MLP. Stochastic gradient descent (SGD) or its variant optimization algorithm, such as the adaptive learning rate Adam algorithm, is used. During the training process, the model parameters are continuously adjusted so that the loss function value of the model on the training set gradually decreases. Through multiple iterations of training, the model converges and achieves a good fitting effect.

[0047] The model optimization submodule within the equipment output prediction module optimizes the selected model. The trained model is evaluated using a test set, and various evaluation metrics are calculated, such as root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). 2 ) etc. RMSE measures the square root of the mean of the sum of squares of the deviations between predicted and actual values, reflecting the accuracy of the prediction; MAE directly calculates the mean of the absolute values ​​of the deviations between predicted and actual values, and is relatively insensitive to outliers; R 2 The RMSE (Real-Time Sequence of Errors) value is used to evaluate the goodness of fit of a model to the data. A value closer to 1 indicates a better fit. For example, if a model predicts the output of a vertical machining center next week, a smaller RMSE value indicates a smaller deviation between the predicted and actual output, and a more accurate prediction.

[0048] Based on the evaluation results, if the model performance does not meet expectations, the model will be optimized. On one hand, the model's hyperparameters will be adjusted, such as increasing the hidden layer depth of the LSTM, expanding the number of nodes in the MLP, and retraining the model. On the other hand, more features may be introduced or feature selection techniques may be used to remove redundant features and improve the model's generalization ability. Simultaneously, if overfitting is detected, regularization techniques, such as L1 and L2 regularization, will be used to constrain model parameters, prevent the model from becoming overly complex, and ensure that the model can still perform well in predicting new data.

[0049] Once the model is optimized, real-time production forecasting and application are performed. The model prediction submodule within the equipment production forecasting module predicts the output of the target equipment. When a new production task is assigned, real-time collected equipment monitoring parameters, equipment output parameters, equipment maintenance parameters, and equipment fault parameters, along with order and process information updated from the Manufacturing Execution System (MES) and Enterprise Resource Planning (ERP) systems, are input into the optimized model to quickly predict the equipment's output under the current operating conditions. For example, if a workshop receives a batch of urgent orders and needs to immediately assess whether the vertical machining center can complete the task on time, the model, combined with real-time data such as the current equipment status and raw material supply, predicts the daily output for the next week, providing a basis for production scheduling decisions.

[0050] Based on production forecasts, production managers can rationally arrange production plans, adjust equipment schedules, and optimize resource allocation. If the predicted output of a certain piece of equipment cannot meet order demand, timely equipment maintenance, increased raw material procurement, or consideration of outsourcing some production tasks can be arranged. If there is a production surplus, equipment maintenance can be arranged in advance to reduce idle time, improve equipment utilization, reduce production costs, and maximize production efficiency.

[0051] The maintenance cycle prediction module mentioned above is used to predict the maintenance cycle of the target equipment based on the current environmental parameters and production conditions of the target equipment, using a pre-trained maintenance cycle prediction model.

[0052] Specifically, the current environmental parameters of the target equipment include, but are not limited to, the ambient temperature and humidity, and dust concentration. Examples of equipment detection parameters have already been given above and will not be repeated here.

[0053] In a specific example, the maintenance cycle prediction module includes: a data processing submodule, a feature extraction submodule, a model training submodule, a model optimization submodule, and a model prediction submodule.

[0054] The data processing submodule within the maintenance cycle prediction module processes the current environmental parameters and equipment monitoring parameters of the target equipment. For example, at the edge computing node, a statistical rule-based method is used to clean the real-time collected equipment operation data. For hydraulic press pressure data, the 3σ criterion is used: if the pressure value at a certain moment exceeds three times the normal mean standard deviation and the duration exceeds a certain threshold (e.g., 10 seconds), it is judged in conjunction with the equipment status log. If it is not caused by process adjustments, it is marked as an outlier and corrected based on normal data from adjacent time periods. Simultaneously, the equipment's built-in self-calibration module automatically calibrates the sensors at fixed intervals to ensure data accuracy.

[0055] The data processing submodule within this maintenance cycle prediction module also processes historical maintenance cycle data. For example, it standardizes the time format for historical maintenance data and extracts key information from unstructured maintenance personnel records using natural language processing technology, transforming it into structured data. For instance, it extracts the fault type as "hydraulic leakage" and the repair measure as "replacing the gasket" from the maintenance personnel's description of "a slight leak was found in the hydraulic press; it returned to normal operation after the gasket was replaced," facilitating subsequent data analysis.

[0056] The feature extraction submodule within the maintenance cycle prediction module extracts features from the processed data. For example, it selects features closely related to the maintenance cycle from the preprocessed data, such as the pressure fluctuation frequency of the hydraulic press, hydraulic oil contamination level (calculated using oil quality sensors or periodic monitoring data), and the cumulative operating time of key components. Min-Max normalization is used to unify feature data of different magnitudes into the 0-1 range, avoiding the impact of data dimension differences on model training performance and ensuring that each feature has equal weight in the model.

[0057] The model training submodule within the maintenance cycle prediction module extracts features for feature training. Based on the characteristics of the equipment data and the maintenance cycle prediction requirements, a combination of Support Vector Machine (SVM) and time series analysis can be used. For equipment with obvious periodic operating characteristics, such as injection molding machines that produce in batches using fixed molds, time series analysis can capture the patterns of equipment state changes within the production cycle. SVM, on the other hand, utilizes its powerful classification capabilities to construct classification boundaries for different maintenance needs based on historical data, mapping the equipment's operating state to the corresponding maintenance cycle category. Furthermore, for equipment data with complex nonlinear relationships, deep learning models such as Recurrent Neural Networks (RNNs) and their variant, Long Short-Term Memory Networks (LSTMs), can be applied. These models can learn the implicit state information during long-term equipment operation and predict maintenance cycles.

[0058] The preprocessed dataset is divided into training and test sets according to a certain ratio, such as selecting 70% of the training set for training and 30% for testing. For the SVM model, a suitable kernel function (such as radial basis function RBF) is selected, and the optimal classification hyperplane is found using an optimization algorithm (such as sequential minimum optimization algorithm SMO) on the training set to determine the model parameters. For RNN or LSTM models, reasonable hyperparameters are set, such as the number of hidden layers and neurons, and stochastic gradient descent (SGD) or Adam optimization algorithm is used. During training, the model parameters are adjusted according to the loss function (such as mean squared error MSE), and the model is converged through multiple iterations to achieve a better fitting effect.

[0059] The model optimization submodule within the maintenance cycle prediction module is used to optimize the selected models. The trained model is evaluated using a test set, with common metrics including accuracy, recall, and F1 score. Accuracy measures the proportion of correctly predicted samples out of the total sample, reflecting the overall prediction precision of the model. Recall indicates the proportion of equipment that actually requires maintenance that is correctly predicted by the model, crucial for avoiding missed maintenance requests. The F1 score comprehensively considers accuracy and recall, balancing the two to provide a more comprehensive evaluation of model performance. For example, if the model predicts that a hydraulic press needs maintenance within the next week, and the actual situation matches, then it is a correct prediction. Various metric values ​​are obtained through statistical analysis of a large number of test samples to evaluate the model's performance.

[0060] Based on the evaluation results, if the model performance does not meet expectations, optimization will be performed. On the one hand, the model hyperparameters will be adjusted, such as the penalty parameter C of SVM, the parameter γ of the RBF kernel function, or the hidden layer depth of RNN. On the other hand, more effective features will be considered, such as combining equipment fault tree analysis to mine potential factors affecting maintenance cycles, removing redundant features, preventing overfitting, improving the model's generalization ability, and ensuring that the model performs well on new data.

[0061] Once the aforementioned model is optimized, maintenance cycles are predicted and applied in real time. The model prediction submodule within the maintenance cycle prediction module is used to predict the maintenance cycle of the target equipment. As the equipment continues to operate, new operational data is collected in real time and fed into the trained and optimized model. Combined with current environmental and production condition information, the next stage of maintenance cycle is quickly predicted. For example, if a stamping press is performing a high-intensity production task, the model, based on real-time collected punch pressure and vibration data, as well as workshop temperature and humidity, predicts that the stamping press requires a comprehensive maintenance check within 3 days after completing the current order, providing the maintenance department with preparation time.

[0062] Based on the maintenance cycle forecast, the maintenance department can rationally arrange maintenance personnel and allocate maintenance resources. If the maintenance cycles of multiple pieces of equipment are predicted to be similar, personnel can be coordinated in advance and corresponding spare parts can be prepared to avoid resource shortages caused by clustered maintenance. For critical equipment, if the maintenance cycle is approaching and production tasks are tight, contingency plans can be developed in advance, such as adjusting production plans and temporarily adding backup equipment, to ensure production continuity, reduce equipment failure rates, and improve overall production efficiency.

[0063] The aforementioned equipment fault prediction module is used to predict the fault information of the target equipment based on the equipment detection parameters and daily operation and maintenance parameters of the target equipment, using a pre-trained equipment fault prediction model.

[0064] Specifically, the equipment detection parameters and daily operation and maintenance parameters of the target equipment have been introduced above and will not be repeated here. This data provides a rich sample basis for building a fault prediction model, helping the model learn the patterns of equipment fault occurrence and their correlation with operation and maintenance.

[0065] In a specific example, the equipment failure prediction module includes: a data processing submodule, a feature extraction submodule, a model training submodule, a model optimization submodule, and a model prediction submodule.

[0066] Specifically, the data processing submodule in the equipment fault prediction module is used to preprocess the equipment detection parameters and daily operation and maintenance parameters of the target equipment. For example, in the data management center, real-time data is cleaned using a combination of statistical rules and machine learning algorithms. For vibration data V, an improved 3σ criterion is adopted, combined with the vibration spectrum characteristics during normal equipment operation, if the vibration amplitude V at a certain moment... t Exceeding the normal mean μv by 3 times the standard deviation σ v And the vibration frequency f t Not within the frequency range corresponding to the normal process [f min ,f max Within this range, it is considered an outlier. That is, when |Vt-uv|>3σ v and At the same time, corrections are made by comparing with historical data from the same period or by using reasonable values ​​predicted by machine learning models. Simultaneously, Kalman filtering is used to remove noise interference from the collected electrical parameters.

[0067] The data processing submodule within the equipment failure prediction module is also used to standardize historical failure and maintenance data, unify the time format, and convert unstructured maintenance records into structured data using natural language processing technology. For example, the record recorded by maintenance personnel, "Abnormal noise was found in the spindle; after disassembly and inspection, the worn bearing was replaced, and the equipment returned to normal operation," is extracted to identify the failure type as "spindle bearing failure" and the maintenance measure as "replacing the bearing." The failure type and maintenance measure are coded to facilitate subsequent model processing.

[0068] Let the set of fault types be T = {T1, T2, ..., T...} n The set of maintenance measures is M = {M1, M2, ..., M}. n The fault type and maintenance measures are coded using one-hot encoding. For fault type T... f If T f =T i Then its encoding vector t f Let M be an n-dimensional vector, where the i-th element is 1 and the rest are 0; similarly, for maintenance measure Mf If M f =M j Then its encoding vector M f Let the vector be an m-dimensional vector, where the j-th element is 1 and the rest are 0. This facilitates subsequent model processing.

[0069] The feature extraction submodule within the equipment fault prediction module is used to extract features closely related to faults from various preprocessed data sets. For example, selecting features closely related to equipment faults from the preprocessed data includes derived features in addition to the directly acquired parameters. For instance, fault characteristic indicators calculated based on the spindle vibration amplitude V and frequency f. in and These are the average vibration amplitude and average frequency during normal operation, based on the motor current change rate. Inferred electrical fault risk coefficient (α is a coefficient) etc.

[0070] The Min-Max normalization or Z-score standardization methods are used to unify the dimensions of feature data of different magnitudes.

[0071] The Min-Max normalization formula is: Where x is the original feature value, x min and x max These are the minimum and maximum values ​​of the feature, x. norm It is the normalized value.

[0072] The Z-score standardization formula is: Where x is the original feature value, μ is the mean of the feature, and σ is the standard deviation of the feature. Normalization ensures that each feature has equal influence during model training, avoiding model bias caused by differences in data scale.

[0073] The model training submodule within the equipment fault prediction module trains the model on features closely related to the extracted faults. For example, an appropriate AI model is selected based on the equipment type, data characteristics, and required fault prediction accuracy. For equipment operation data with significant time-series characteristics, such as long-term accumulated hourly operating status monitoring data, Long Short-Term Memory (LSTM) networks and their variant, Gated Recurrent Units (GRUs), are ideal choices. LSTM networks are combined with Autoregressive Moving Average (ARIMA) models for fault prediction. Taking motor equipment as an example, the ARIMA model is first used to perform preliminary analysis on the motor's current, temperature, and other time-series data, capturing seasonality and trend characteristics, and extracting residual sequences. These residual sequences are then input into the LSTM model along with the original data. Leveraging its ability to learn long-term dependencies, the LSTM model delves deeper into the hidden fault precursors within the data. For instance, if the motor's operating current exhibits a small but continuous upward trend over a long period, and the LSTM model detects an abnormal correlation between the current change rate and temperature rise, it can predict 4-6 hours in advance that the motor may fail due to winding overheating, allowing for timely repairs.

[0074] The model optimization submodule within the equipment failure prediction module optimizes the selected models. For example, the preprocessed dataset is divided into training and test sets according to a certain ratio, such as selecting 70% of the training set for training and 30% of the test set for testing. The trained model is then evaluated using the test set, with evaluation metrics including accuracy, recall, F1 score, and root mean square error (RMSE).

[0075] Accuracy reflects the proportion of samples correctly predicted by the model out of the total sample. The formula is: Where TP is the true positive (the number of samples that the model predicts are positive and are actually positive), TN is the true negative (the number of samples that the model predicts are negative and are actually negative), FP is the false positive (the number of samples that the model predicts are positive but are actually negative), and FN is the false negative (the number of samples that the model predicts are negative but are actually positive).

[0076] Recall measures the proportion of actual faulty samples that are accurately predicted by the model.

[0077] F1 score: Taking into account both precision and recall, and balancing the relationship between the two, the calculation formula is as follows: in For accuracy.

[0078] Root Mean Square Error (RMSE): Primarily used to measure the square root of the mean of the sum of squares of the deviations between predicted and actual values. It intuitively reflects the accuracy of the prediction and is especially suitable for the prediction and evaluation of continuous variables.

[0079] For example, if a model predicts that a CNC machine tool will experience a spindle failure within the next 24 hours, and the actual situation matches this prediction, then the prediction is correct. Various indicator values ​​are obtained through statistical analysis of a large number of test samples to accurately evaluate the model's performance. Based on the evaluation results, if the model's performance does not meet expectations, targeted optimization work is carried out.

[0080] Hyperparameter tuning: Fine-tune the model's hyperparameters, such as increasing the hidden layer depth of the LSTM, expanding the number of nodes in the DNN, and optimizing the kernel parameters of the SVM, and then retrain the model. Grid search can be used to find the optimal combination of hyperparameters. Grid search performs a comprehensive search on a predefined hyperparameter grid, evaluates the model's performance on the validation set for each hyperparameter combination, and selects the combination with the best performance.

[0081] Feature Optimization: Deeply explore more potential effective features, combining methods such as equipment fault tree analysis and expert experience to expand the feature space. Fault tree analysis identifies more potential factors that may lead to equipment failure and incorporates them as new features into the model. Simultaneously, feature selection techniques, such as analysis of variance (ANOVA) and correlation analysis, are used to eliminate redundant features, effectively preventing overfitting. For ANOVA, the variance of each feature is calculated; if the variance of a feature is too small, it indicates that the feature's value does not vary significantly and contributes little to the model's discriminative ability, and it can be considered for removal. This comprehensive approach enhances the model's generalization ability, ensuring that the model maintains good predictive performance when faced with new data.

[0082] Once the above model is optimized, it is used for real-time production forecasting and application. The model prediction submodule within the equipment failure prediction module is used to predict equipment failures of the target equipment. As the target equipment continues to operate, new real-time data is continuously input into the optimized and trained model. Combined with current environmental and production condition information, the model quickly predicts whether the equipment has a failure risk in the future. Let the input real-time feature vector be x. new The model output is the failure probability P. fault The calculation method differs for different models. For example, for a binary SVM model, the result can be further transformed into a probability form after calculation using the decision function; for an LSTM model, the output layer is processed by the Softmax function to obtain the probability distribution of each category, where the probability of the fault category is given by the Softmax function.

[0083] Once the failure probability P is predicted fault Exceeding the preset threshold P t Immediately issue warnings to operations and maintenance personnel through multiple channels. The warning information includes a fault type prediction (T). pred Possible time t pred Severity of the fault faultDegree and preliminary investigation recommendations A suggest For example, if it is predicted that there is a 70% probability that the spindle bearing of a CNC machine tool will fail within the next 12 hours, it is recommended that maintenance personnel stop the machine in advance to check the spindle bearing temperature and vibration, and prepare the tools and spare parts needed to replace the bearing.

[0084] Based on the fault prediction results, the operation and maintenance department can formulate scientific and reasonable response strategies in advance.

[0085] Preventative maintenance and production plan adjustments: For equipment nearing failure, preventative maintenance should be scheduled promptly if production schedules permit. Let W be the remaining workload for the current production task of the equipment. remain The estimated time required to complete this task is t. task The predicted failure time is t. pred If t task <t pred Equipment shutdown for maintenance can be scheduled after the current task is completed; if t task >t pred If so, production plans need to be adjusted, such as transferring some production tasks to standby equipment, in order to reduce the impact of equipment downtime on production.

[0086] For temporary allocation and emergency repair arrangements, equipment can be temporarily allocated if production tasks are tight. The available time of the equipment is t. available t should be ensured available <t pred To ensure production continuity, emergency repair teams are organized to conduct efficient repairs as soon as equipment malfunctions occur. The repair team can prepare repair tools and spare parts in advance based on the fault type predictions and preliminary troubleshooting suggestions in the early warning information, thus shortening repair time.

[0087] For operation and maintenance plan optimization: Based on long-term failure prediction data, optimize equipment operation and maintenance plans. For example, by analyzing the failure patterns of different equipment under different operating conditions, rationally arrange the equipment maintenance cycle Tc. If a certain piece of equipment has a high failure frequency under high-intensity production conditions, its maintenance cycle can be appropriately shortened; for equipment under low-intensity production conditions, the maintenance cycle can be appropriately extended. At the same time, based on the components that frequently fail in failure predictions, rationally adjust the spare parts procurement plan to ensure that spare parts inventory can meet maintenance needs without causing inventory backlog, thereby reducing equipment failure rates from the source, improving overall production efficiency, and ensuring the stable and smooth operation of enterprise production.

[0088] The equipment spare parts prediction module mentioned above is used to predict the spare parts demand of the target equipment based on the equipment testing parameters, daily operation and maintenance parameters, production task parameters and current environmental parameters of the target equipment, using a pre-trained equipment spare parts prediction model.

[0089] Specifically, equipment testing parameters, daily operation and maintenance parameters, and current environmental parameters have been illustrated above and will not be repeated here. Production task parameters include, but are not limited to, equipment type, equipment batch, equipment process complexity, and daily output requirements.

[0090] In a specific example, the equipment spare parts prediction module includes: a data processing submodule, a feature extraction submodule, a model training submodule, a model optimization submodule, and a model prediction submodule.

[0091] The data processing submodule within the equipment spare parts prediction module cleans real-time collected equipment operation data at the edge computing node using a statistical rule-based method. For temperature data from vertical machining center motors, the 3σ criterion is applied: if the temperature value at a certain moment exceeds three times the normal mean standard deviation and lasts for more than a certain threshold (e.g., 5 minutes), it is judged in conjunction with the equipment status log. If it is not caused by process adjustments, it is marked as an outlier and corrected based on normal data from adjacent time periods. Simultaneously, the equipment's built-in self-calibration module automatically calibrates sensors at fixed intervals to ensure data accuracy.

[0092] The data processing submodule within the equipment spare parts prediction module is also used to standardize and organize historical fault and maintenance data, converting unstructured maintenance personnel records into structured data using natural language processing technology. For example, the maintenance personnel's record, "A jamming phenomenon was found when the spindle box of the vertical machining center moved. After disassembly and inspection, it was found that the ball screw nut pair was severely worn. After replacing the worn ball screw nut pair, the equipment returned to normal operation," extracts the fault location as "spindle box," the fault cause as "wear of the ball screw nut pair," and the maintenance measure as "replacing the ball screw nut pair." Key information such as the type of spare parts is also encoded to facilitate subsequent model processing.

[0093] The feature extraction submodule within the equipment spare parts prediction module is used to extract features closely related to spare parts from various preprocessed data. In addition to the directly collected parameters, derived features are also selected from the preprocessed data that are closely related to spare parts demand.

[0094] Component loss rate R wear Based on the equipment's running time T run and fault frequency f fault Calculate using the formula R. wear =α*T run +β*f fault , where α and β are weighting coefficients, which can be obtained through statistical analysis of historical data.

[0095] Spare parts consumption coefficient C consume Based on production task intensity I taskEstimate, assuming the production task intensity and daily output Q daily Complexity of the process (C) complex Related, can be represented as I task =γ*Q daily +δ*C complex (γ and δ are weighting coefficients), then C consume =η*I task (η is the proportionality coefficient).

[0096] The Min-Max normalization or Z-score standardization methods are used to unify the dimensions of feature data of different magnitudes.

[0097] The Min-Max normalization formula is: Where x is the original feature value, x min and x max These are the minimum and maximum values ​​of the feature, x. norm It is the normalized value.

[0098] The Z-score standardization formula is: z = (x - μ) / σ, where x is the original feature value, μ is the mean of the feature, and σ is the standard deviation of the feature. Normalization ensures that each feature has equal influence during model training, avoiding model bias caused by differences in data scale.

[0099] The model training submodule within the equipment spare parts prediction module trains the model on features closely related to spare parts after feature extraction. Based on the characteristics of the equipment data and the needs for spare parts prediction, time series prediction models such as ARIMA (Autoregressive Moving Average) and its extended form SARIMA (Seasonal Autoregressive Moving Average) can be selected. For equipment with obvious seasonal production patterns or periodic maintenance needs, these models can effectively capture the time-series trends in the data and predict the demand cycle and quantity changes of spare parts.

[0100] The ARIMA model can be represented as ARIMA(p,d,q), where p is the autoregressive order, d is the differencing order, and q is the moving average order. Its formula is: φ(B)(1-B) d Y t =θ(B)∈ t

[0101] Where Yt is the time series data, B is the lag operator, and φ(B) = 1 - φ1B - ... - φ p B p It is an autoregressive coefficient polynomial, θ(B)=1+θ1B+…+θ q B q It is a moving average coefficient polynomial, ε t It is a white noise sequence.

[0102] The SARIMA model, based on ARIMA, considers seasonality and can be expressed as SARIMA(p,d,q)(P,D,Q), where P, D, and Q are the orders of the seasonal autoregressive, differencing, and moving average, respectively, and s is the seasonal period. Its formula is:

[0103] φ(B)Φ(B s (1-B) d (1-B s ) D Y t =θ(B)Θ(B) s )∈ t

[0104] Wherein, Φ(B s ) and Θ(B s ) is a seasonal autoregressive and moving average coefficient polynomial.

[0105] In addition, machine learning models such as random forests and gradient boosting trees (GBM) also perform well. Random forests, through the integration of multiple decision trees, can uncover the potential correlation between spare parts demand and various factors from complex multi-source data. GBM, on the other hand, is based on the gradient boosting strategy to gradually optimize the decision tree model, thereby improving prediction accuracy. It is especially suitable for processing data with strong nonlinear relationships.

[0106] The preprocessed dataset is divided into a test set and a test set according to a certain ratio, such as selecting 70% of the training set for training and selecting 30% of the test set for testing.

[0107] For the selected model, set appropriate hyperparameters. For example, for the ARIMA model, set the autoregression order, moving average order, etc.

[0108] For random forests, settings such as the number of trees and the maximum tree depth are configured. When each decision tree splits at a node, the optimal feature is selected based on the information gain formula (IG(S,A)=H(S)-H(S|A)), where S is the sample set and A is the feature. It is entropy, p i It is the probability of category i.

[0109] For GBM, set the learning rate α, the number of iterations M, etc.

[0110] Stochastic Gradient Descent (SGD) or its variant Adam algorithm are employed. During training, the goal is to minimize the loss function; for regression problems like spare parts demand prediction, the mean squared error loss function is commonly used. Through multiple iterations of training, the model parameters are continuously adjusted until the model converges, achieving a good fit.

[0111] The model optimization submodule within the spare parts prediction module optimizes the selected models. For example, it evaluates the trained models using a test set, with common metrics including root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE).

[0112] Root Mean Square Error (RMSE): Measures the square root of the mean of the sum of squares of the deviations between predicted and actual values, reflecting the accuracy of the prediction.

[0113] Mean Absolute Error (MAE): The mean of the absolute values ​​of the deviations between the predicted and actual values ​​is calculated directly, and it is relatively insensitive to outliers.

[0114] Mean Absolute Percentage Error (MAPE): Presents prediction error as a percentage, making it easy to intuitively compare the prediction performance of different models or datasets.

[0115] For example, if the model predicts that a vertical machining center will need to replace 5 sets of ball screw nut pairs within the next month, and 6 sets are actually replaced, the model's performance can be accurately evaluated by statistically analyzing a large number of test samples to obtain various indicator values.

[0116] Based on the evaluation results, if the model performance does not meet expectations, optimization will be carried out. On the one hand, the model hyperparameters will be adjusted, such as increasing the number of trees in the random forest and reducing the learning rate of GBM, and the model will be retrained. On the other hand, more effective features will be introduced, and the factors that may affect the demand for spare parts will be mined by combining equipment fault tree analysis, removing redundant features, preventing overfitting, improving the model's generalization ability, and ensuring that the model performs well on new data.

[0117] Based on the evaluation results, if the model performance does not meet expectations, optimization will be carried out. On the one hand, the model hyperparameters will be adjusted, such as increasing the number of trees in the random forest and reducing the learning rate of GBM, and the model will be retrained. On the other hand, more effective features will be introduced, and the factors that may affect the demand for spare parts will be mined by combining equipment fault tree analysis, removing redundant features, preventing overfitting, improving the model's generalization ability, and ensuring that the model performs well on new data.

[0118] Once the aforementioned model is optimized, real-time prediction and application of equipment spare parts are performed. The model prediction submodule within the equipment spare parts prediction module described above is used to predict the spare parts required for the target equipment. As the target equipment continues to operate, new operational data is collected in real time and fed into the trained and optimized model. Combined with current environmental and production condition information, the model quickly predicts the types, quantities, and demand timing of spare parts required by the equipment in the near future. For example, if a vertical machining center is performing high-intensity machining tasks, the model, based on real-time collected spindle temperature, tool wear data, workshop temperature and humidity, and production schedule, predicts that three sets of tool holders and two coolant pump seals will need to be replaced within the next two weeks, providing the purchasing department with advance preparation time.

[0119] Based on spare parts forecasts, the purchasing department can develop reasonable procurement plans, optimize purchase orders according to demand timelines and quantities, and avoid inventory backlogs or stockouts. For critical and fragile spare parts, long-term partnerships should be established with suppliers in advance to ensure timely supply. Meanwhile, the warehousing department can adjust inventory layout based on forecasts, placing frequently used spare parts in easily accessible locations to improve replenishment efficiency, ensure smooth equipment operation and maintenance, reduce equipment downtime due to spare parts shortages, and improve overall production efficiency.

[0120] The aforementioned equipment energy consumption prediction module is used to predict the energy consumption parameters of the target equipment based on the equipment detection parameters, current environmental parameters, and daily energy consumption parameters of the target equipment, using a pre-trained equipment energy consumption prediction model.

[0121] Specifically, the equipment detection parameters and current environmental parameters have been illustrated above and will not be repeated here. Daily energy consumption parameters include, but are not limited to, daily, weekly, and monthly power consumption.

[0122] In a specific example, the device energy consumption prediction module includes: a data processing submodule, a feature extraction submodule, a model training submodule, a model optimization submodule, and a model prediction submodule.

[0123] The data processing submodule within the equipment energy consumption prediction module processes the target equipment's detection parameters, current environmental parameters, and daily energy consumption parameters. For example, at edge computing nodes, a statistical rule-based method is used to clean the real-time collected equipment operation data. For power data, the 3σ criterion is employed: if the power value at a certain moment exceeds three times the normal mean standard deviation and lasts for more than a certain threshold (e.g., 10 seconds), it is judged in conjunction with the equipment status log. If it is not caused by normal power fluctuations (e.g., the moment the compressor starts), it is marked as an outlier and corrected based on normal data from adjacent time periods. Simultaneously, sensors are periodically calibrated using a standard source to ensure data accuracy; for example, a high-precision power source is used to calibrate the power sensor to ensure that the power measurement error is within the allowable range.

[0124] The 3σ criterion assumes the mean of the power data is μ. P The standard deviation is σ p If the power value P at a certain moment t Satisfying |Pt-μ P |>3σ p And the duration t continue >t th (t th If a threshold (e.g., 10 seconds) is used, it is marked as an outlier.

[0125] Standard deviation calculation: Where P i Let be the i-th power data point, and n be the number of data points.

[0126] Mean calculation:

[0127] The data processing submodule within the aforementioned equipment energy consumption prediction module is also used to standardize the format of historical energy consumption data. For example, it resamples energy consumption data at equal time intervals (e.g., per hour) to fill in missing values. Min-Max normalization or Z-score standardization methods are employed to unify the dimensions of equipment operating parameters, environmental parameters, and energy consumption data at different scales, avoiding the impact of data scale differences on model training performance. For instance, spindle power, table displacement, and ambient temperature data are normalized to the 0-1 range, ensuring that each feature has equal weight in model training.

[0128] Min-Max normalization: For data x, the normalized value x norm The calculation formula is: Where x min and x max These are the minimum and maximum values ​​of the data, respectively.

[0129] Z-score standardization: z = (x - μ) / σ, where x is the original feature value, μ is the mean of the feature, and σ is the standard deviation of the feature.

[0130] The feature extraction submodule within the equipment energy consumption prediction module trains the model on features closely related to equipment energy consumption after feature extraction. For example, features closely related to energy consumption and energy efficiency ratio are extracted from the preprocessed data. In addition to the directly collected parameters, derived features are also included, such as cutting power calculated based on spindle speed, feed rate, and cutting force, and average energy consumption rate calculated based on equipment runtime and power consumption. These derived features can more intuitively reflect the energy consumption characteristics of the equipment and provide strong support for model prediction.

[0131] Cutting power calculation: Pc = Fc * Vc, where Pc is the cutting power, Fc is the cutting force, and Vc is the cutting speed (which can be calculated from the spindle speed and tool diameter).

[0132] Average energy consumption rate calculation: Venergy=E / t, where Venergy is the average energy consumption rate, E is the power consumption, and t is the equipment operating time.

[0133] The model training submodule within the equipment energy consumption prediction module trains the extracted features closely related to equipment energy consumption. Based on the characteristics of the equipment data and the energy consumption prediction requirements, a multiple linear regression model can be selected. For equipment where energy consumption has an approximately linear relationship with multiple operating parameters and environmental factors, such as some simple auxiliary equipment, it can quickly establish linear correlations between variables and predict energy consumption. For vertical machining centers with complex nonlinear relationships, deep learning models such as Long Short-Term Memory (LSTM) networks and their variant Gated Recurrent Unit (GRU) are better choices. They can effectively capture the dynamic changes in equipment operating status over time and the implicit nonlinear relationships between energy consumption and various factors. Furthermore, Support Vector Regression (SVR) performs well in small-sample, high-dimensional data scenarios, accurately predicting energy consumption and energy efficiency ratio by constructing the optimal regression hyperplane.

[0134] The preprocessed dataset is divided into a training set and a test set according to a certain ratio, such as selecting 70% of the training set for training and 30% of the test set for testing. For the selected model, reasonable hyperparameters are set, such as the number of hidden layers and neurons in the LSTM, and the kernel function parameters in the SVR. Stochastic gradient descent (SGD) or its variants, such as the Adam algorithm with adaptive learning rate, are used to minimize the loss function (e.g., mean squared error, mean absolute error) during training. The model parameters are continuously adjusted, and training is iterated multiple times until the model converges and achieves a good fit.

[0135] Among them, the model optimization submodule in the equipment energy consumption prediction module trains the features closely related to equipment energy consumption after feature extraction.

[0136] The trained model is evaluated using a test set. Common metrics include root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). 2 ) etc. RMSE measures the square root of the mean of the sum of squares of the deviations between predicted and actual values, reflecting the accuracy of the prediction; MAE directly calculates the mean of the absolute values ​​of the deviations between predicted and actual values, and is relatively insensitive to outliers; R 2 Used to evaluate the goodness of fit of a model to data, the closer to 1, the better the fit. For example, if the model predicts that the energy consumption of a vertical machining center will be 500 kWh tomorrow, and the actual energy consumption is 520 kWh, the model's performance can be accurately evaluated by statistically analyzing a large number of test samples to obtain the values ​​of various indicators.

[0137] Based on the evaluation results, if the model performance does not meet expectations, optimization will be carried out. On the one hand, the model hyperparameters will be adjusted, such as increasing the hidden layer depth of LSTM and optimizing the kernel function parameters of SVR, and the model will be retrained. On the other hand, more effective features will be introduced, and the potential factors affecting energy consumption and energy efficiency ratio will be explored by combining the energy balance principle of equipment and thermodynamic knowledge. Redundant features will be removed to prevent overfitting, improve the model's generalization ability, and ensure that the model performs well on new data.

[0138] Once the above model is optimized, the energy consumption of the equipment is predicted and applied in real time. The model prediction submodule in the equipment energy consumption prediction module is used to predict the energy consumption of the target equipment.

[0139] As equipment continues to operate, new operational data is collected in real time and fed into a trained and optimized model. Combined with current environmental and production condition information, the model quickly predicts the equipment's energy consumption and energy efficiency ratio over a future period. For example, when a vertical machining center in a factory is performing high-intensity machining tasks, the model predicts, based on real-time collected data such as spindle power, feed axis displacement, and workshop temperature and humidity, that energy consumption will continue to rise and the energy efficiency ratio will gradually decrease over the next four hours, providing an early warning to the energy management department.

[0140] Based on energy consumption and energy efficiency ratio forecasts, energy management departments can rationally formulate energy dispatch plans. If a significant increase in energy consumption is predicted, power distribution can be adjusted in advance to avoid overload tripping. For periods with low energy efficiency ratios, equipment operating parameters can be optimized, such as adjusting spindle speed and feed rate, to improve the energy efficiency ratio and reduce energy costs. Simultaneously, based on long-term forecast data, the effectiveness of equipment energy-saving retrofits can be evaluated, providing data support for equipment upgrades and achieving the company's energy conservation and emission reduction goals.

[0141] Energy dispatch strategy: Establishing an energy cost function Where P i It is the predicted power for the i-th time period, t i c is the duration of the i-th time interval. iLet be the electricity price for the i-th time period. Through an optimization algorithm, the operating time and power allocation of equipment are adjusted to minimize the energy cost function C while meeting production demands. For example, during periods with lower electricity prices, the operating time or load rate of equipment is increased; during periods with higher electricity prices, the operating intensity of equipment is appropriately reduced or some non-critical tasks are suspended.

[0142] Equipment operating parameter optimization: Based on the relationship model between energy efficiency ratio and equipment operating parameters, optimization algorithms such as gradient descent are used to find the optimal combination of operating parameters. For example, for the two parameters spindle speed n and feed rate v, the energy efficiency ratio EER can be expressed as f(n,v). By calculating the gradient of EER with respect to n and v, the values ​​of n and v are continuously adjusted to maximize EER. In actual operation, after each parameter adjustment, it is necessary to monitor the operational stability and processing quality of the equipment to ensure that the parameter adjustment does not negatively impact production.

[0143] Energy-saving retrofit effectiveness evaluation: Before and after the energy-saving retrofit of the equipment, collect energy consumption and production data for a period of time (e.g., one month before and one month after the retrofit). Compare the average energy consumption E before and after the retrofit. before and E after and energy consumption per unit output E unit-before and E unit-after This is used to evaluate the effectiveness of energy-saving renovations. The energy saving rate η can be calculated as η = (E before -E after ) / E unit-before *100%. Simultaneously, the changes in energy efficiency ratio before and after the renovation were analyzed, and the overall effectiveness of the energy-saving renovation in improving equipment performance and energy utilization efficiency was comprehensively evaluated. Based on the evaluation results, lessons learned were summarized to provide a reference for subsequent energy-saving renovation projects.

[0144] In some optional implementations, the IoT device prediction system in this disclosure further includes:

[0145] The data acquisition module is used to collect equipment testing parameters, equipment output parameters, equipment operation and maintenance parameters, and equipment fault parameters of the target equipment.

[0146] The data acquisition module also includes various testing instruments or sensors used for testing equipment, such as temperature sensors, displacement sensors, vibration sensors, and current sensors.

[0147] The data acquisition module also includes various testing instruments or sensors for detecting equipment output parameters.

[0148] The data acquisition module also includes various testing instruments or sensors for detecting equipment operation and maintenance parameters.

[0149] The data acquisition module also includes various testing instruments or sensors for detecting equipment fault parameters.

[0150] In some optional implementations, the IoT device prediction system in this disclosure further includes:

[0151] The equipment data display module is used to display the production output forecast, maintenance cycle forecast, fault information pre-handling results, spare parts demand forecast, and energy consumption parameter forecast of the target equipment.

[0152] The equipment data display module intuitively displays real-time production output forecasts, maintenance cycle forecasts, fault information pre-handling results, spare parts demand forecasts, and energy consumption parameter forecasts for the target equipment. In addition, this module can also display equipment operating data. Key parameters for various equipment, such as temperature, noise, pressure, current, and speed, are displayed in real-time using charts and figures. Users can view the current operating status and working environment of the target equipment, as well as the production output forecasts, maintenance cycle forecasts, fault information pre-handling results, spare parts demand forecasts, and energy consumption parameter forecasts, anytime via computers, mobile phones, or other terminal devices. For example, production managers can use the real-time data display module to understand the operating status of each piece of equipment on the production line and adjust production plans accordingly. Maintenance personnel can use the operating parameters from the equipment data display module to handle equipment malfunctions promptly. The equipment data display module provides users with timely and accurate equipment operating information, helping to improve the efficiency of equipment management and the scientific basis of decision-making.

[0153] In some optional implementations, the IoT device prediction system in this disclosure further includes:

[0154] The equipment fault alarm module is used to issue fault alarms to the target equipment based on the predicted fault information.

[0155] For example, based on the predicted fault information of the target equipment, the module can issue fault alarms. For instance, when equipment malfunctions, its operating parameters become abnormal, or a safety hazard arises, the fault alarm module will issue a real-time alarm. Alarm methods can include various forms such as audible alarms, SMS alarms, and email alarms. Users can set alarm thresholds and alarm methods according to their needs. For example, when the pressure of a hydraulic press exceeds a set safety value, the fault alarm module will immediately issue an audible alarm and send an SMS notification to relevant personnel. The fault alarm module ensures that equipment faults and safety issues are addressed promptly, protecting the safety of equipment and personnel. Simultaneously, the module records alarm history, facilitating user review and analysis of equipment operation.

[0156] In some optional implementations, the IoT device prediction system in this disclosure further includes:

[0157] The equipment safety execution module is used to implement safety measures when a target device has a safety hazard. This module establishes different safety indicator systems based on different equipment types. For example, for lifting equipment such as overhead cranes and elevators, safety indicators include lifting capacity, lifting height, operating speed, and braking performance; for high-pressure equipment such as vacuum furnaces, safety indicators include pressure, temperature, and electrical insulation performance; for machining equipment such as CNC machine tools and milling machines, safety indicators include spindle speed, cutting force, and tool condition.

[0158] By monitoring the operating parameters of the target equipment in real time and comparing them with corresponding safety indicators, it is determined whether there are any operational risks associated with the target equipment. If the operating parameters of the target equipment are detected to exceed the safety indicator range, the equipment safety execution module will issue an alarm. For example, if the lifting capacity of the crane exceeds the rated value, or if the operating speed of the elevator increases abnormally, an alarm will be issued immediately to remind operators and managers to take appropriate measures.

[0159] For equipment posing a risk, the system can decelerate and stop for maintenance based on pre-set control measures, ensuring safety. For example, for a hydraulic press, if the pressure rises abnormally, the system can automatically reduce the hydraulic system pressure or stop the equipment to prevent malfunctions or accidents. Simultaneously, the safety management module records safety events related to the equipment for subsequent analysis and improvement of safety management measures.

[0160] The equipment safety execution module provides strong protection for the safety of target equipment and personnel by establishing a scientific safety indicator system and early warning and prediction mechanism, ensuring that the factory's production activities can proceed safely and stably.

[0161] In some optional implementations, the IoT device prediction system in this disclosure further includes:

[0162] The device interface management module is used to match the target interface and target interface protocol according to the device type of the target device.

[0163] The device interface management module is responsible for managing the interfaces of external devices. For various types of equipment within the factory, such as CNC machine tools and vertical machining centers, data acquisition and transmission are achieved through corresponding interface protocols. Simultaneously, the system can be integrated with other enterprise management systems, such as ERP and MES systems, to achieve information sharing and collaboration. The interface management module ensures the system's compatibility and scalability, providing strong technical support for intelligent equipment management.

[0164] In some optional implementations, the IoT device prediction system in this disclosure embodiment further includes: a user management module for managing user permission information and user personal information.

[0165] Users can register for device management through the system by filling in personal information and requesting permissions. After the administrator approves the registration, the user can log in to the system using their registered account and password.

[0166] User management devices allow for different user permissions based on user roles and responsibilities. For example, administrators have the highest privileges and can manage the system comprehensively; device managers can view device operating status and create maintenance plans; and maintenance personnel can receive fault alarms and view fault diagnosis results.

[0167] User management devices can manage user information, including modifying user information, deleting user accounts, and resetting user passwords. Users can also modify their personal information and passwords themselves.

[0168] In some optional implementations, the IoT device prediction system in this disclosure embodiment further includes: a device resource management module for managing device resources.

[0169] The equipment resource management module is used to manage basic equipment information, such as equipment name, model, and manufacturer. This module also allows for the entry, modification, and deletion of equipment information. Users can quickly query equipment resources and understand the overall status of the equipment through this module.

[0170] In some optional implementations, the IoT device prediction system in this disclosure embodiment further includes: a device performance evaluation module, used to evaluate the overall performance of the target device based on the production output prediction results, maintenance cycle prediction results, fault information prediction results, spare parts demand prediction results, and energy consumption parameter prediction results of the target device.

[0171] The equipment performance evaluation module can assess the overall performance of the target equipment based on various prediction results. Simultaneously, the module allows for the customization of equipment performance indicators, real-time calculation of equipment performance levels, and analysis of performance trends by comparing performance with historical data, thus helping to optimize equipment operating efficiency. This deep integration of data analysis, machine learning, and equipment management gives edge computing units a significant advantage in the field of intelligent equipment management.

[0172] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An Internet of Things (IoT) device prediction system, characterized in that, include: The equipment output prediction module is used to predict the production output of the target equipment based on the equipment detection parameters, equipment output parameters, daily operation and maintenance parameters and equipment fault parameters of the target equipment, using a pre-trained output prediction model. The maintenance cycle prediction module is used to predict the maintenance cycle of the target equipment based on the current environmental parameters and equipment detection parameters of the target equipment, using a pre-trained maintenance cycle prediction model. The equipment fault prediction module is used to predict the fault information of the target equipment based on the equipment detection parameters and daily operation and maintenance parameters of the target equipment, using a pre-trained equipment fault prediction model. The equipment spare parts prediction module is used to predict the spare parts demand of the target equipment based on the equipment testing parameters, daily operation and maintenance parameters, production task parameters and current environmental parameters of the target equipment, using a pre-trained equipment spare parts prediction model. The equipment energy consumption prediction module is used to predict the energy consumption parameters of the target equipment based on the equipment detection parameters, current environmental parameters, and daily energy consumption parameters of the target equipment, using a pre-trained equipment energy consumption prediction model.

2. The IoT device prediction system according to claim 1, characterized in that, The equipment output prediction module, the maintenance cycle prediction module, the equipment failure prediction module, the equipment spare parts prediction module, and the equipment energy consumption prediction module all include corresponding data processing submodules, feature extraction submodules, model training submodules, model optimization submodules, and model prediction submodules.

3. The IoT device prediction system according to claim 1, characterized in that, Also includes: The data acquisition module is used to collect the equipment detection parameters, equipment output parameters, daily operation and maintenance parameters, and equipment fault parameters of the target equipment.

4. The IoT device prediction system according to claim 1, characterized in that, Also includes: The equipment data display module is used to display the production output forecast, maintenance cycle forecast, fault information pre-handling results, spare parts demand forecast, and energy consumption parameter forecast of the target equipment.

5. The IoT device prediction system according to claim 1, characterized in that, Also includes: The equipment fault alarm module is used to issue a fault alarm to the target equipment based on the fault information prediction results of the target equipment.

6. The IoT device prediction system according to claim 1, characterized in that, Also includes: The device safety execution module is used to execute safety measures when the target device has safety hazards.

7. The IoT device prediction system according to claim 1, characterized in that, Also includes: The device interface management module is used to match the target interface and the target interface protocol according to the device type of the target device.

8. The IoT device prediction system according to claim 1, characterized in that, Also includes: The user management module is used to manage user permission information and user personal information.

9. The IoT device prediction system according to claim 1, characterized in that, Also includes: The equipment resource management module is used to manage equipment resources.

10. The IoT device prediction system according to claim 1, characterized in that, Also includes: The equipment performance evaluation module is used to evaluate the overall performance of the target equipment based on the production output prediction results, maintenance cycle prediction results, fault information prediction results, spare parts demand prediction results, and energy consumption parameter prediction results of the target equipment.

Images