A method and system for monitoring intelligent factory equipment based on the Internet of Things

Abstract

The application discloses a kind of based on the equipment health state monitoring method and system of Internet of Things.The method is by obtaining the data in multiple source sensors and pre-processing, obtains the real-time running dataset of equipment, determines health quantitative deviation value and health state trend by real-time running dataset.When health state is abnormal, the starting point coordinate of abnormal event is recorded and marked abnormal event trigger timestamp, and the dependence relationship between devices is analyzed using graph network modeling.Furthermore, the potential risk probability is analyzed by long short-term memory network.The application realizes real-time monitoring of equipment health state, fault propagation prediction and risk assessment, and improves the operation reliability and maintenance efficiency of industrial equipment.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a smart factory equipment monitoring method and system based on IoT. Background Technology

[0002] Driven by Industry 4.0, the deep integration of IoT technology and manufacturing is reshaping production paradigms. Smart factory equipment monitoring, as a core support link in the intelligent manufacturing system, directly impacts production efficiency optimization, equipment lifecycle management, and enterprise operational risk control capabilities. Currently, mainstream equipment monitoring systems generally adopt discrete, periodic data collection models and are mostly based on isolated, single-dimensional indicator analysis frameworks. This makes it difficult to construct a continuous health profile of equipment operating status, let alone analyze the implicit process coupling relationships and fault propagation paths in multi-device collaborative operations. This limitation leads to enterprises often only being able to passively respond to abnormal situations in equipment health management, lacking proactive risk prediction capabilities, resulting in downtime losses or increased maintenance costs. Summary of the Invention

[0003] The technical solution adopted by this invention to solve its technical problem is: a smart factory equipment monitoring method based on the Internet of Things, comprising the following steps:

[0004] S1: Obtain multi-source sensor data from the device from the IoT framework, and preprocess the data in the multi-source sensor data to obtain the real-time operating dataset of the device;

[0005] S2: Using the real-time running dataset, determine the health quantification deviation value and health status trend value of the device. If the health quantification deviation value exceeds a preset threshold, it is recorded as an abnormal event of the device. At the same time, the starting coordinates of the abnormal event are recorded and the trigger timestamp of the abnormal event is marked.

[0006] S3: Obtain the timestamp of the abnormal event and the real-time running dataset of the device, and analyze the adjacency distance value and association weight value of the device through graph network modeling technology to obtain the dependency matrix between the devices;

[0007] S4: Use a long short-term memory network to fuse the health status trend value with the timestamp of the abnormal event to determine the probability of the occurrence of potential risks.

[0008] Furthermore, in the IoT-based smart factory equipment monitoring method described in this invention, the data in the multi-source sensor data includes: abnormal temperature fluctuation values, vibration peak frequency numbers, and current abrupt change time points collected through data cleaning technology.

[0009] Furthermore, in the IoT-based smart factory equipment monitoring method of the present invention, determining the health status trend value of the equipment through the real-time operation dataset includes: calculating the health quantification deviation value using a sliding window algorithm for the real-time operation dataset, and determining the health status trend value of the equipment based on a weighted combination of the abnormal temperature fluctuation value, the vibration peak frequency number, and the current sudden change time point.

[0010] Furthermore, the smart factory equipment monitoring method based on the Internet of Things described in this invention also includes step S301 after step S3:

[0011] For the dependency matrix, a propagation analysis algorithm is used to calculate the propagation path length. Based on the starting coordinates of the abnormal event, the correlation weight value, and the weight decay coefficient, the fault propagation speed value and path priority ranking are determined.

[0012] Furthermore, the smart factory equipment monitoring method based on the Internet of Things described in this invention also includes step S302 after step S3:

[0013] The initial influence radius value is extracted from the path priority ranking, and the starting coordinates of the abnormal event and the fault propagation speed value are analyzed by Bayesian network to obtain the hierarchical response probability value.

[0014] Furthermore, the IoT-based smart factory equipment monitoring method of the present invention further includes the following steps:

[0015] S5: Based on the probability of occurrence of the potential risks, a situation map is generated using dynamic visualization technology. A dynamic monitoring view of the equipment is obtained by comprehensively mapping and adjusting the mapping weight based on the health quantification deviation value, the transmission path length value, and the graded response probability value.

[0016] Furthermore, the IoT-based smart factory equipment monitoring method of the present invention further includes the following steps:

[0017] S6: Extract high-risk areas from the dynamic monitoring view whose initial influence radius value exceeds the preset range, and adjust the sensor data acquisition frequency within the device's adjacent distance value through the feedback control algorithm to obtain the updated real-time operation dataset.

[0018] Furthermore, in the IoT-based smart factory equipment monitoring method of the present invention, determining the health status trend value of the equipment through the real-time operating dataset includes the following steps:

[0019] S201: Extract time series features from the real-time running dataset using the sliding window algorithm to obtain the feature extraction set;

[0020] S202: Based on the feature extraction set, a weighted combination method is used to fuse data on temperature anomalies, vibration peaks, and current mutations to obtain a fused feature set;

[0021] S203: Calculate the distribution of deviation values ​​from the fused feature set, and use mean smoothing technique to process the noise in the distribution to obtain a smoothed deviation set;

[0022] S204: For the smoothing deviation set, if the deviation value exceeds the preset threshold, the mutation location is determined by time point analysis to obtain the mutation location set;

[0023] S205: Based on the mutation localization set, time series analysis is used to obtain the dynamic changes in health status, resulting in a set of status changes;

[0024] S206: Calculate the trend value distribution of health quantification using the set of state changes to obtain the trend distribution set;

[0025] S207: Use a linear regression algorithm to predict future changes in the health status of equipment from the trend distribution set to obtain a predicted trend set.

[0026] Furthermore, in the trusted verification strategy method for portal websites described in this invention, step S6 specifically includes the following steps:

[0027] S601: Initial image data is extracted through dynamic monitoring to obtain the impact distribution set;

[0028] S602: Calculate the influence radius value based on the influence distribution set to obtain the radius feature set;

[0029] S603: If the radius feature concentration value exceeds the preset range, then a high-risk area is determined and a risk distribution set is obtained;

[0030] S604: Use a feedback control algorithm to adjust the adjacency distance parameter for the risk distribution set to obtain the distance adjustment set;

[0031] S605: Update the sensor data acquisition frequency by adjusting the distance set to obtain the frequency distribution set;

[0032] S606: Obtain real-time running data based on the frequency distribution set to obtain the running update set;

[0033] S607: When comparing the running update set with the preset range, if the deviation exceeds the threshold, the decision tree algorithm is used to optimize and adjust the direction to obtain the direction set.

[0034] Furthermore, the present invention also provides an Internet of Things-based smart factory equipment monitoring system for executing the monitoring method described in the above technical solution.

[0035] The present invention offers the following advantages: It discloses a method and system for monitoring the health status of equipment based on the Internet of Things (IoT). This method preprocesses data from multiple sensors to obtain a real-time operational dataset of the equipment. The dataset is then used to determine the quantitative deviation value and trend of the health status. When the health status is abnormal, the starting coordinates of the abnormal event are recorded and the trigger timestamp is marked. Graph network modeling is used to analyze the dependencies between devices. Furthermore, by analyzing the probability of potential risks using a Long Short-Term Memory (LSTM) network, this invention achieves real-time monitoring of equipment health status, fault propagation prediction, and risk assessment, thereby improving the operational reliability and maintenance efficiency of industrial equipment. Attached Figure Description

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0037] Figure 1 This is a schematic diagram illustrating the steps of a device health status monitoring method based on the Internet of Things according to the present invention;

[0038] Figure 2 This is a schematic diagram of steps S5 and S6 of a device health status monitoring method based on the Internet of Things according to the present invention.

[0039] Figure 3 This is a schematic diagram illustrating the specific steps of step S2 in the IoT-based device health status monitoring method of the present invention.

[0040] Figure 4 This is a schematic diagram illustrating the specific steps of step S6 in the IoT-based device health status monitoring method of the present invention. Detailed Implementation

[0041] To provide a clearer understanding of the technical features, objectives, and effects of this invention, specific embodiments are now described in detail with reference to the accompanying drawings. In the following description, it should be understood that the orientations or positional relationships indicated by terms such as "front," "rear," "upper," "lower," "left," "right," "longitudinal," "horizontal," "vertical," "horizontal," "top," "bottom," "inner," "outer," "head," and "tail" are based on the orientations or positional relationships shown in the accompanying drawings, and are constructed and operated in a specific orientation. They are only for the convenience of describing this technical solution and do not indicate that the device or element referred to must have a specific orientation; therefore, they should not be construed as limitations on this invention.

[0042] It should also be noted that, unless otherwise explicitly specified and limited, terms such as "installation," "connection," "linking," "fixing," and "setting" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. When an component is referred to as being "on" or "below" another component, the component can be located "directly" or "indirectly" on the other component, or there may be one or more intermediary components. The terms "first," "second," "third," etc., are only for the convenience of describing this technical solution and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0043] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0044] like Figures 1 to 4 As shown, in the first embodiment of the present invention, a smart factory equipment monitoring method based on the Internet of Things is provided, including the following steps (S1, S2, S3, S4, S5 and S6):

[0045] Step S1: Obtain multi-source sensor data from the device from the IoT framework, and preprocess the data in the multi-source sensor data to obtain the device's real-time operating dataset.

[0046] Specifically, the data from multi-source sensors includes: abnormal temperature fluctuations, vibration peak frequencies, and current abrupt change times collected through data cleaning techniques. It should be noted that the data acquired from multi-source sensors through the IoT framework can be diverse, including temperature, vibration, and current data. In industrial equipment, abnormal temperature fluctuations, vibration peak frequencies, and current abrupt change times are preferred as core monitoring indicators. These indicators can reflect the equipment's health status, process stability, and potential risks in real time from multiple dimensions, providing crucial support for predictive maintenance, energy efficiency optimization, and safe production.

[0047] Furthermore, temperature, vibration, and current data are acquired from multiple sensors using an IoT framework. Data cleaning techniques are employed to preprocess abnormal temperature fluctuations, vibration peak frequencies, and current abrupt change time points to obtain an initial operational dataset. If the abnormal temperature fluctuations in the initial operational dataset exceed a preset threshold, a sliding window technique is used to calculate the fluctuation trend, resulting in a temperature change feature set. Based on this feature set, the K-means algorithm is used to cluster the fluctuation trends, determining the abnormal temperature distribution pattern. The vibration peak frequencies are compared with a preset frequency threshold; if the peak frequencies exceed the threshold, the frequency change rate is calculated, resulting in a vibration anomaly index set. The temporal correlation between the vibration anomaly index set and the current abrupt change time points is obtained, and time series analysis is used to identify abrupt changes in the equipment's operating state, resulting in a state abrupt change dataset. For this state abrupt change dataset, mean smoothing is used to process noise between abrupt change points, resulting in a smoothed equipment operating trend set. Based on this smoothed set, a linear regression algorithm is used to predict the changing trend of the equipment's operating state, resulting in a predicted state dataset.

[0048] For example, when acquiring multi-source sensor data from an IoT framework, the first step is to process the collected abnormal temperature fluctuations using data cleaning techniques. For instance, a sliding window algorithm is used to smooth the temperature data with a window size of 5, removing abnormal fluctuations and retaining data within the normal range. Next, the peak vibration frequencies are analyzed. A Fast Fourier Transform (FFT) algorithm is used to convert the time-domain signal to the frequency-domain signal, extracting the main frequency components. For example, a peak frequency of 50Hz is identified as a normal vibration frequency of the device. Then, current abrupt changes are detected. A differential algorithm is used to calculate the rate of change of current, with a threshold of 0.5A / s. When the rate of change exceeds the threshold, the abrupt change time point is recorded. For example, a current abrupt change at 10:05:23 is marked as an abnormal event. Finally, the cleaned temperature, vibration, and current data are integrated into a real-time operating dataset for the device. Timestamp alignment ensures data consistency; for example, the temperature, vibration frequency, and current value at 10:05:23 are stored together to form a complete operating status record. The entire process is implemented through automated scripts, ensuring efficient and accurate data processing.

[0049] Step S2: By running the dataset in real time, determine the health quantification deviation value and health status trend value of the device. If the health quantification deviation value exceeds the preset threshold, it is recorded as an abnormal event of the device. At the same time, the starting coordinates of the abnormal event are recorded and the abnormal event trigger timestamp is marked.

[0050] Specifically, determining the health status trend value of the equipment by using a real-time running dataset includes: calculating the health quantification deviation value using a sliding window algorithm for the real-time running dataset; and determining the health status trend value of the equipment based on a weighted combination of abnormal temperature fluctuations, vibration peak frequencies, and current abrupt change time points.

[0051] In a preferred embodiment, such as Figure 3 As shown, determining the health status trend value of the device by running the dataset in real time includes the following steps (S201, S202, S203, S204, S205, S206, and S207):

[0052] S201: Extract time series features from the real-time running dataset using the sliding window algorithm to obtain the feature extraction set;

[0053] S202: Based on the feature extraction set, a weighted combination method is used to fuse data on temperature anomalies, vibration peaks, and current mutations to obtain a fused feature set;

[0054] S203: Calculate the distribution of deviation values ​​from the fused feature set, and use mean smoothing technique to process the noise in the distribution to obtain a smoothed deviation set;

[0055] S204: For the smoothing deviation set, if the deviation value exceeds the preset threshold, the mutation location is determined by time point analysis to obtain the mutation location set;

[0056] S205: Based on the mutation localization set, time series analysis is used to obtain the dynamic changes in health status, resulting in a set of status changes;

[0057] S206: Calculate the trend value distribution of health quantification using the set of state changes to obtain the trend distribution set;

[0058] S207: Use a linear regression algorithm to predict future changes in the health status of equipment from the trend distribution set to obtain a predicted trend set.

[0059] For example, based on a real-time running dataset, a sliding window with a window length of 10 is used to calculate the health quantification deviation value. The standard deviation of the temperature data within the window is calculated, and if it exceeds a threshold of 2.5℃, it is marked as abnormal fluctuation. At the same time, the energy proportion of the 80-120Hz frequency band in the vibration spectrum is extracted. When the proportion exceeds 15%, it is judged as a high-frequency anomaly. Combined with abrupt events with a current change rate exceeding 1.2A / s, weight coefficients of 0.4, 0.3, and 0.3 are assigned respectively. The health status trend value is calculated using the linear weighting formula H=0.4×ΔT+0.3×Ef+0.3×ΔI, where ΔT is the normalized value of the temperature standard deviation, Ef is the proportion of high-frequency energy, and ΔI is the number of current abrupt changes. An alert is triggered when the H value exceeds 0.7 for three consecutive windows. For example, if a device calculates ΔT=0.8 (corresponding to a standard deviation of 3.2℃), Ef=0.4 (18% high-frequency energy), and ΔI=0.9 (three sudden changes within 5 minutes) at 14:32:17, and the weighted H=0.73, the system automatically generates a level three health alert and pushes it to the operation and maintenance platform. This process is handled in real time by a streaming computing engine, updating the health status matrix every 30 seconds, and persistently storing historical trend values ​​for degradation analysis.

[0060] S3: Obtain the timestamp of the abnormal event trigger and the real-time running dataset of the device. Analyze the device adjacency distance value and the association weight value through graph network modeling technology to obtain the dependency matrix between devices.

[0061] Specifically, the process involves acquiring event trigger timestamps and real-time operational data from associated devices. Graph network modeling techniques are used to calculate adjacency distances and association weights between devices, resulting in a device dependency matrix. Adjacency distances and association weights are extracted from the dependency matrix, and matrix factorization is used to analyze the dependency strength between devices, yielding a dependency strength set. For this dependency strength set, clustering algorithms are employed to divide the devices into association groups, resulting in a device association grouping set. The fluctuation characteristics of real-time operational data within each group are obtained from these grouping sets, resulting in a fluctuation feature set. If outliers exceeding a preset threshold exist in the fluctuation feature set, timestamps corresponding to the outliers are marked using timestamp location techniques, resulting in an outlier timestamp set. Based on the outlier timestamp set, statistical analysis methods are used to calculate the dynamic change probability of device dependencies in each group, resulting in a probability distribution set. Significantly changed dependencies are extracted from the probability distribution set, and the device dependency matrix is ​​adjusted using graph network update techniques, resulting in an updated dependency matrix.

[0062] For example, after obtaining the event trigger timestamp, an adjacency graph between devices is constructed using graph network modeling techniques, combined with the real-time operational dataset of the associated devices. The Dijkstra algorithm is used to calculate the adjacency distance between devices. The initial distance between device A and device B is set to 10, and the distance between device B and device C is set to 15. The shortest path calculation yields an adjacency distance of 25 between device A and device C. Based on the operational data between devices, the PageRank algorithm is used to calculate the association weights. The initial weights for device A and device B are set to 0.3, for device B to 0.5, and for device C to 0.2. Through iterative calculation, the final weights for device A are 0.35, for device B to 0.45, and for device C to 0.20. The adjacency distances and weights are combined to construct a device dependency matrix. In this matrix, the dependency value between device A and device B is 0.8, the dependency value between device B and device C is 0.7, and the dependency value between device A and device C is 0.6. Matrix analysis identified device B as a core node, with two devices having a dependency value exceeding 0.7, indicating its high influence within the network. Further K-means clustering was used to group the dependency matrix, setting the cluster number k=2. Devices A and C were grouped together, while device B was grouped separately, resulting in the device dependency grouping results. Based on these grouping results and device operational data, a device dependency report was generated for subsequent device optimization and maintenance strategy development.

[0063] In a preferred embodiment, such as Figure 2 As shown, after step S3, the following step S301 is also included: For the dependency matrix, the propagation analysis algorithm is used to calculate the propagation path length value, and the fault propagation speed value and path priority ranking are determined based on the coordinates of the abnormal event starting point, the correlation weight value and the weight decay coefficient.

[0064] Specifically, propagation path data is obtained through a dependency matrix, and the length of each path is calculated using a propagation analysis algorithm to obtain a path length set. Based on the path length set and the coordinates of the starting point of the abnormal event, the initial propagation path for the abnormality is determined, resulting in an initial path set. For the initial path set, the association weight values ​​are obtained, and the weight data is adjusted using a weight decay coefficient to obtain an adjusted weight set. If there are weight values ​​in the adjusted weight set that exceed a preset threshold, the fault propagation speed is calculated using a propagation analysis algorithm to obtain a speed distribution set. Based on the speed distribution set and the path length set, the path priority is determined to obtain a priority sequence set. The propagation path order is updated using the priority sequence set, and the path data is adjusted using a dependency matrix to obtain an updated path set. By updating the path set and the speed distribution set, the dynamic trend of fault propagation is obtained, resulting in a trend set.

[0065] For example, based on the device dependency matrix, the random walk model in the propagation analysis algorithm is used to calculate the fault propagation path length. The starting point of the abnormal event is set as device D, with an initial weight of 0.4. The weight of the adjacent device E is 0.6, and the weight of device F is 0.3. According to the weight decay coefficient of 0.2, the propagation path length from device D to device E is calculated as 1 / (0.6 × 0.8) = 2.08, and the propagation path length from device D to device F is 1 / (0.3 × 0.8) = 4.17. By traversing all device nodes using the Floyd-Warshall algorithm, the complete propagation path matrix is ​​obtained. The shortest path from device D to device E is 2.08, and the path length from device E to device G is 3.33. Therefore, the total propagation path length from device D to device G is 5.41. Based on the fault propagation delay parameter in real-time operational data, the propagation speed from device D to device E is 2.08 / 0.5 = 4.16 units of time, and the propagation speed from device D to device G is 5.41 / 0.5 = 10.82 units of time. According to the path length and propagation speed values, a priority queue sorting algorithm is used to generate a path priority list. A threshold of 5.0 is set, marking the path from device D to device E as high priority and the path from device D to device G as low priority. Through propagation path analysis, device E is identified as a critical propagation node, with its impact covering three downstream devices. Based on the priority sorting results and combined with device operating status data, a fault propagation prediction report is generated for preventative maintenance decision-making.

[0066] Furthermore, such as Figure 2 As shown, step S3 is followed by the following step S302: extracting the initial influence radius value from the path priority sorting, and analyzing the starting coordinates of the abnormal event and the fault propagation speed value through a Bayesian network to obtain the graded response probability value.

[0067] Specifically, the initial impact radius value is obtained by prioritizing paths to determine the range data for triggering abnormal events, resulting in a range distribution set. A Bayesian network is used to jointly analyze the range distribution set and the starting coordinates to obtain the conditional probability of propagation speed, resulting in a probability distribution set. Based on the probability distribution set and the dynamic characteristics of fault propagation, the triggering conditions for tiered responses are determined, resulting in a condition set. If there are probability values ​​in the condition set exceeding a preset threshold, the tiered response order is adjusted based on the propagation speed, resulting in an adjustment sequence set. Based on the adjustment sequence set and the impact radius, the expansion trend of abnormal events is obtained, resulting in a trend distribution set. Using the trend distribution set and response probability values, the priority processing direction of fault propagation is determined, resulting in a direction set. Based on the direction set and the condition set, a decision tree algorithm is used to adjust the update order of path priorities, resulting in an update distribution set.

[0068] For example, in fault propagation analysis, the initial influence radius can be determined by calculating the weighted average distance between devices surrounding the origin of the abnormal event. Assuming the origin of the abnormal event is device A, and its influence radius includes devices B (distance 1.5), C (distance 2.3), and D (distance 3.1), with weights of 0.7, 0.5, and 0.3 respectively, then the initial influence radius is (1.5×0.7+2.3×0.5+3.1×0.3) / (0.7+0.5+0.3)=2.12. A fault propagation probability model is constructed based on a Bayesian network. The coordinates of the origin of the abnormal event (x=3, y=5) and the propagation speed (e.g., 4.2 units of time from device A to device B) are input, and the graded response probability is calculated using a conditional probability table.

[0069] For example, the response probability is 0.9 when the propagation velocity is below 3.0; 0.6 when it is between 3.0 and 6.0; and 0.2 when it is above 6.0. Combining the real-time status data of device A (such as a temperature deviation of 2.8℃ and vibration amplitude of 0.15g), a fault propagation path is simulated using Markov chain Monte Carlo sampling. The output response probabilities are 0.73 for device B, 0.58 for device C, and 0.31 for device D. Based on a probability threshold of 0.5, an early warning signal for device B is automatically triggered, and a response strategy matrix containing the key node (device B) and potentially affected devices (devices C and D) is generated.

[0070] S4: Use a long short-term memory network to fuse the health status trend value with the timestamps of abnormal events to determine the probability of potential risks occurring.

[0071] Specifically, by using tiered response probability values, a long short-term memory (LSTM) network is employed to fuse health status trend values ​​and event trigger timestamps to obtain a distribution set of potential risk occurrence probabilities. For this distribution set and timestamps, the changing trends of potential risks are obtained, resulting in a trend distribution set. If any probability of occurrence in the trend distribution set exceeds a preset threshold, the weights of the fusion analysis are adjusted using the LSM network, resulting in an adjusted probability set. Based on the adjusted probability set and health status trend values, the priority order of potential risks is determined, resulting in a priority sequence set. Using the priority sequence set and event trigger timestamps, the dynamic distribution of risk occurrence is obtained, resulting in a dynamic distribution set. For the dynamic distribution set and tiered responses, a decision tree algorithm is used to adjust the update order of probability values, resulting in an updated distribution set. Based on the updated distribution set and the trend distribution set, the processing direction for potential risks is determined, resulting in a direction set.

[0072] For example, in an equipment health management system, a Long Short-Term Memory (LSTM) network is used to fuse and analyze the equipment health status trend values ​​with event trigger timestamps to determine the probability of potential risks occurring. Assume that the health status trend values ​​of equipment E include the temperature change rate (0.05℃ / min), vibration frequency change rate (0.02Hz / min), and current fluctuation rate (0.1A / min). These data are input into the LSTM network, which contains 128 hidden units and is trained using the Adam optimizer with a learning rate of 0.001. The event trigger timestamps include the equipment E's startup time (2023-10-01 08:00:00) and the time of its most recent maintenance (2023-09-25 10:00:00). These timestamps are encoded as time-series features and input into the LSTM network along with the health status trend values. Through forward propagation calculation using the LSTM network, the probability of potential risks occurring for equipment E within the next 24 hours is calculated to be 0.68. Based on historical failure data of equipment E (e.g., 3 failures in the past year, with an average failure interval of 120 days), a Poisson distribution model is used to model the failure frequency, further refining the risk probability to 0.72. According to the risk probability threshold of 0.7, the system automatically generates a maintenance plan for equipment E, including a suggested maintenance time (2023-10-05 14:00:00) and required maintenance resources (e.g., bearing replacement, sensor calibration). Simultaneously, the system pushes the risk information for equipment E to relevant management personnel and generates a response strategy matrix containing maintenance task allocation and resource scheduling.

[0073] Furthermore, such as Figure 2 As shown, the IoT-based smart factory equipment monitoring method of the present invention further includes the following steps:

[0074] S5: By using dynamic visualization technology to generate a situation map based on the probability of potential risks, and by using a comprehensive mapping and mapping weight adjustment value based on the health quantification deviation value, transmission path length value, and graded response probability value, a dynamic monitoring view of the equipment is obtained.

[0075] By using occurrence probability and graded response, a situation map is generated using dynamic visualization technology, resulting in an initial view set. Based on the initial view set and health quantification, deviation values ​​are calculated, yielding a deviation distribution set. For the deviation distribution set and transmission paths, length values ​​are obtained, resulting in a path feature set. Through the path feature set and comprehensive mapping, mapping weights are adjusted, resulting in a weight distribution set. Based on the weight distribution set and dynamic adjustments, a monitoring view is generated, resulting in a view update set. For the view update set and situation map, if the deviation value exceeds a preset threshold, a decision tree algorithm is used to update the weight distribution set, resulting in an optimized distribution set. Through the optimized distribution set and the monitoring view, the direction of dynamic adjustment is determined, resulting in a direction set. In IoT smart factories, the dynamic monitoring view of equipment can visualize real-time data, historical trends, and multi-dimensional indicators, providing intuitive support for equipment management, fault early warning, and process optimization.

[0076] For example, based on multi-dimensional data fusion analysis of equipment health status, a dynamic heatmap algorithm is used to spatially map the health quantification deviation values. Temperature deviation is weighted at 0.4, vibration deviation at 0.3, and current deviation at 0.3. Local density values ​​are calculated using a Gaussian kernel function to generate a thermal distribution map of the equipment surface. Areas with abnormal temperatures (such as bearing locations) are displayed in red (RGB value 255,0,0). An alarm is triggered when the deviation value exceeds a threshold of 1.5. The conduction path analysis uses a graph neural network (GNN) for modeling. Nodes represent sensors (e.g., T1-T6), and edge weights are determined by the signal correlation coefficient (0.2-0.8). The shortest conduction path is calculated using Dijkstra's algorithm (e.g., T3→T5→T6, total length 2.1). Paths exceeding a set value of 3.0 are marked as critical paths. The graded response probability is dynamically updated using a Bayesian network. The prior probability is set at 0.6. When the real-time vibration amplitude exceeds 4.5 mm / s, the likelihood function is adjusted to 1.8, and the posterior probability increases to 0.82. The above data is used to synthesize a 3D situation map in real time through the WebGL rendering engine. The viewpoint is automatically adjusted to follow the device topology. When the health status change rate exceeds 0.2 / s, dynamic focusing is activated. At the same time, a JSON format monitoring report containing 12 key parameters is generated and updated to the STATUS_MAP field of the database table TB_MONITOR_VIEW every 5 seconds.

[0077] Furthermore, such as Figure 2 As shown, the IoT-based smart factory equipment monitoring method of the present invention further includes the following steps:

[0078] S6: Extract high-risk areas from the dynamic monitoring view whose initial influence radius value exceeds the preset range, and adjust the sensor data acquisition frequency within the device's adjacent distance value through the feedback control algorithm to obtain the updated real-time operation dataset.

[0079] Specifically, such as Figure 4 As shown, in the trusted verification strategy method for portal websites described in this invention, step S6 specifically includes the following steps (S601, S602, S603, S604, S605, S606 and S607):

[0080] S601: Extract initial impact data through dynamic monitoring to obtain the impact distribution set;

[0081] S602: Calculate the influence radius value based on the influence distribution set to obtain the radius feature set;

[0082] S603: If the radius feature concentration value exceeds the preset range, then a high-risk area is determined and a risk distribution set is obtained;

[0083] S604: Use a feedback control algorithm to adjust the adjacency distance parameter for the risk distribution set to obtain the distance adjustment set;

[0084] S605: Update the sensor data acquisition frequency by adjusting the distance set to obtain the frequency distribution set;

[0085] S606: Obtain real-time running data based on the frequency distribution set to obtain the running update set;

[0086] S607: When comparing the running update set with the preset range, if the deviation exceeds the threshold, the decision tree algorithm is used to optimize and adjust the direction to obtain the direction set.

[0087] For example, when the initial influence radius of a certain area in the dynamic monitoring view exceeds the preset range (e.g., a set threshold of 2.5 meters, but an actual detected radius of 3.8 meters), the system automatically triggers a neighborhood sensor network reconstruction mechanism. A Kalman filter-based feedback control algorithm is used, with the center point of the abnormal area (coordinates X=4.2, Y=7.6) as the reference, to dynamically adjust the sampling frequency of 12 vibration sensors (V1-V12) within a 5-meter radius. Specifically, the sampling rate of V3, V7, and V9 within 2 meters of the center point is increased from 100Hz to 500Hz, while V2, V5, and V8 between 2 and 4 meters are increased to 300Hz, and the rest maintain their original frequencies. In the data fusion stage, a wavelet transform algorithm is used to perform time-frequency analysis on the multi-source signals. The db4 wavelet basis function is selected for 5-level decomposition, and the energy characteristics of the 1.6-3.2kHz frequency band are extracted as the main analysis indicators. For the reconstructed dataset, the root mean square (RMS) value of each sensor node is calculated using a sliding window (window width 2 seconds, step size 0.5 seconds). When the RMS value of node V7 exceeds 1.8 times the baseline value for three consecutive windows (baseline value 4.3 mm / s, detected value 7.9 mm / s), the device topology adaptive module is activated, reducing the linkage coefficient between adjacent motor units M4 and M5 from 0.7 to 0.4. Simultaneously, the opening of the cooling system flow valve is increased by 15% using a PID controller (proportional coefficient Kp=1.2, integral time Ti=8s). The updated operating data is compressed and stored in a time-series database. The Delta coding algorithm is used to compress the original data volume by 62%. A real-time data snapshot containing 23-dimensional feature vectors including spatial coordinates, frequency domain features, and control parameters is generated every 10 seconds and written to the / real_time / path of the distributed storage cluster. The file naming rule is TIMESTAMP_DEVICEID_CRC32 checksum format. In IoT-enabled smart factories, extracting high-risk areas through dynamic monitoring and dynamically adjusting the sampling frequency of adjacent sensors can significantly improve system responsiveness, optimize resource utilization efficiency, and provide higher-quality data support for real-time decision-making. Furthermore, subsequent analysis and decision-making based on the updated real-time operational dataset (high-frequency, high-quality, risk-focused) can significantly enhance the system's intelligence level and production efficiency.

[0088] This invention discloses a device health status monitoring method based on the Internet of Things (IoT). The method acquires data from multiple sensors, preprocesses abnormal data such as temperature, vibration, and current, and uses a sliding window algorithm to calculate the health quantification deviation value to determine the trend of device health status. When the health status is abnormal, graph network modeling is used to analyze the dependencies between devices, and a propagation analysis algorithm is used to determine the fault propagation path. Furthermore, Bayesian networks and long short-term memory networks are used to analyze the probability of potential risks and generate a dynamic situation map. Finally, for high-risk areas, the data acquisition frequency of relevant sensors is adjusted. This invention achieves real-time monitoring of device health status, fault propagation prediction, and risk assessment, improving the operational reliability and maintenance efficiency of industrial equipment.

[0089] In a second embodiment of the present invention, an Internet of Things-based smart factory equipment monitoring system is provided for performing the monitoring method as described in the first embodiment.

[0090] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0091] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0092] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0093] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They do not limit the scope of protection of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A smart factory equipment monitoring method based on the Internet of Things, characterized in that, Includes the following steps: S1: Obtain multi-source sensor data from the device from the IoT framework, and preprocess the data in the multi-source sensor data to obtain the real-time operating dataset of the device; S2: Using the real-time running dataset, determine the health quantification deviation value and health status trend value of the device. If the health quantification deviation value exceeds a preset threshold, it is recorded as an abnormal event of the device. At the same time, the starting coordinates of the abnormal event are recorded and the trigger timestamp of the abnormal event is marked. S3: Obtain the timestamp of the abnormal event and the real-time operation dataset of the device, and analyze the device adjacency distance value and association weight value through graph network modeling technology to obtain the dependency matrix between the devices; S4: Use a long short-term memory network to fuse the health status trend value with the timestamp of the abnormal event to determine the probability of the occurrence of potential risks; Following step S3, step S301 is also included: For the dependency matrix, a propagation analysis algorithm is used to calculate the propagation path length. Based on the starting coordinates of the abnormal event, the correlation weight value, and the weight decay coefficient, the fault propagation speed value and path priority ranking are determined.

2. The method for monitoring smart factory equipment based on the Internet of Things according to claim 1, characterized in that, The data in the multi-source sensor data includes: abnormal temperature fluctuation values, vibration peak frequency numbers, and current abrupt change time points collected through data cleaning technology.

3. The method for monitoring smart factory equipment based on the Internet of Things according to claim 2, characterized in that, The step of determining the health status trend value of the device using the real-time running dataset includes: calculating the health quantification deviation value using a sliding window algorithm for the real-time running dataset, and determining the health status trend value of the device based on a weighted combination of the abnormal temperature fluctuation value, the number of vibration peak frequencies, and the time points of the current sudden change.

4. The method for monitoring smart factory equipment based on the Internet of Things according to claim 1, characterized in that, It also includes step S302 following step S3: The initial influence radius value is extracted from the path priority ranking, and the starting coordinates of the abnormal event and the fault propagation speed value are analyzed by Bayesian network to obtain the hierarchical response probability value.

5. The method for monitoring smart factory equipment based on the Internet of Things according to claim 4, characterized in that, It also includes the following steps: S5: Based on the probability of occurrence of the potential risks, a situation map is generated using dynamic visualization technology. A dynamic monitoring view of the equipment is obtained by comprehensively mapping and adjusting the mapping weight based on the health quantification deviation value, the transmission path length value, and the graded response probability value.

6. The method for monitoring smart factory equipment based on the Internet of Things according to claim 5, characterized in that, It also includes the following steps: S6: Extract high-risk areas from the dynamic monitoring view whose initial influence radius value exceeds the preset range, and adjust the sensor data acquisition frequency within the device's adjacent distance value through the feedback control algorithm to obtain the updated real-time operation dataset.

7. The method for monitoring smart factory equipment based on the Internet of Things according to claim 1, characterized in that, Determining the health status trend value of the device using the real-time running dataset includes the following steps: S201: Extract time series features from the real-time running dataset using the sliding window algorithm to obtain the feature extraction set; S202: Based on the feature extraction set, a weighted combination method is used to fuse data on temperature anomalies, vibration peaks, and current mutations to obtain a fused feature set; S203: Calculate the distribution of deviation values ​​from the fused feature set, and use mean smoothing technique to process the noise in the distribution to obtain a smoothed deviation set; S204: For the smoothing deviation set, if the deviation value exceeds the preset threshold, the mutation location is determined by time point analysis to obtain the mutation location set; S205: Based on the mutation localization set, time series analysis is used to obtain the dynamic changes in health status, resulting in a set of status changes; S206: Calculate the trend value distribution of health quantification using the set of state changes to obtain the trend distribution set; S207: Use a linear regression algorithm to predict future changes in the health status of equipment from the trend distribution set to obtain a predicted trend set.

8. A smart factory equipment monitoring method based on the Internet of Things according to claim 6, characterized in that, Step S6 specifically includes the following steps: S601: Extract initial impact data through dynamic monitoring to obtain the impact distribution set; S602: Calculate the influence radius value based on the influence distribution set to obtain the radius feature set; S603: If the radius feature concentration value exceeds the preset range, then a high-risk area is determined and a risk distribution set is obtained; S604: Use a feedback control algorithm to adjust the adjacency distance parameter for the risk distribution set to obtain the distance adjustment set; S605: Update the sensor data acquisition frequency by adjusting the distance set to obtain the frequency distribution set; S606: Obtain real-time running data based on the frequency distribution set to obtain the running update set; S607: When comparing the running update set with the preset range, if the deviation exceeds the threshold, the decision tree algorithm is used to optimize and adjust the direction to obtain the direction set.

9. A smart factory equipment monitoring system based on the Internet of Things, characterized in that, Used to perform the IoT-based smart factory equipment monitoring method as described in any one of claims 1 to 8.

Images