An Internet of Things-based fault monitoring method and system for a belt tying machine

Abstract

The present application relates to the technical field of data processing, and discloses a method and system for monitoring the faults of a banding machine based on the Internet of Things. The method includes: collecting multi-source operation data through sensors; performing filtering and segmentation processing on the data to obtain feature data; analyzing the feature data to obtain abnormal identifiers and fault types; establishing a component association map to locate the root cause of the fault and predict the development trend; calculating the maintenance priority to generate a maintenance plan; recording the actual maintenance results and the prediction deviation to update the fault library and knowledge parameters. Through multi-source data fusion, intelligent fault diagnosis, predictive maintenance, and knowledge self-adaptive update mechanism, the present application realizes the early identification and accurate positioning of the faults of the banding machine, effectively reduces the unplanned downtime, and improves the equipment reliability and production efficiency.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a method and system for monitoring the faults of a strapping machine based on the Internet of Things. Background Art

[0002] A strapping machine is a commonly used packaging device in industrial production, mainly used for bundling and fixing products or components, and is widely used in fields such as logistics, manufacturing, and food processing. With the improvement of the level of industrial automation, modern strapping machines have developed into complex devices integrating machinery, electricity, and control, and their operating status directly affects production efficiency and product quality. Traditional maintenance of strapping machines mainly relies on two modes: regular maintenance and post-fault repair, that is, inspection and maintenance are carried out according to a fixed cycle, or maintenance is carried out only after obvious faults occur in the equipment. Some advanced enterprises have introduced simple condition monitoring technologies, and key component status is monitored in real time by installing a single type of sensor (such as a vibration or temperature sensor), and an alarm is triggered when the monitoring data exceeds a preset threshold.

[0003] However, the existing methods for monitoring the faults of strapping machines have obvious deficiencies. First, the traditional regular maintenance method lacks pertinence, which may not only lead to waste of maintenance resources but also cannot effectively prevent sudden faults. Second, the monitoring method using a single sensor cannot comprehensively reflect the complex working state of the equipment, resulting in low fault diagnosis accuracy and high false alarm and missed alarm rates. Third, the existing monitoring systems generally lack intelligent analysis capabilities and cannot predict fault trends based on historical data, making it difficult to achieve early identification and prevention of faults. Finally, the existing methods lack a closed-loop learning mechanism and cannot automatically optimize the diagnosis model according to actual maintenance results, and it is difficult to continuously improve the diagnosis accuracy. These deficiencies lead to low maintenance efficiency of strapping machines, frequent unplanned shutdowns, and seriously affect production continuity and the economic benefits of enterprises. Summary of the Invention

[0004] This application provides a method and system for monitoring the faults of a strapping machine based on the Internet of Things, which is used to realize early identification and accurate positioning of strapping machine faults through multi-source data fusion, intelligent fault diagnosis, predictive maintenance, and a knowledge self-adaptive update mechanism, effectively reduce unplanned shutdown time, and improve equipment reliability and production efficiency.

[0005] In a first aspect, the present application provides a method for monitoring the faults of a strapping machine based on the Internet of Things. The method for monitoring the faults of the strapping machine based on the Internet of Things includes: collecting the operation data of each key part of the strapping machine through a sensor network to obtain multi-source operation data including motor temperature data, tension control parameters, displacement data of the bundling robotic arm, operation status of the cutting device, and strapping deformation rate; filtering and segmenting the data according to the multi-source operation data to obtain the state characteristic data of the strapping machine with time stamps; analyzing and comparing the operation status of the strapping machine based on the state characteristic data of the strapping machine to obtain an abnormal state identifier of the strapping machine and a determination result of the fault type; establishing an associated graph of the strapping machine components according to the abnormal state identifier of the strapping machine and the determination result of the fault type to obtain a fault cause location result and a predicted value of the fault development trend; calculating the maintenance priority in combination with the strapping machine maintenance database according to the fault cause location result and the predicted value of the fault development trend to obtain a maintenance plan and an operation time suggestion; recording the actual maintenance result and the predicted deviation according to the operation effect of the strapping machine after the implementation of the maintenance plan to obtain an updated fault feature library of the strapping machine and maintenance knowledge parameters.

[0006] In a second aspect, the present application provides a system for monitoring the faults of a strapping machine based on the Internet of Things. The system for monitoring the faults of the strapping machine based on the Internet of Things includes:

[0007] An extraction module, configured to collect the operation data of each key part of the strapping machine through a sensor network to obtain multi-source operation data including motor temperature data, tension control parameters, displacement data of the bundling robotic arm, operation status of the cutting device, and strapping deformation rate;

[0008] A segmentation module, configured to filter and segment the data according to the multi-source operation data to obtain the state characteristic data of the strapping machine with time stamps;

[0009] A comparison module, configured to analyze and compare the operation status of the strapping machine based on the state characteristic data of the strapping machine to obtain an abnormal state identifier of the strapping machine and a determination result of the fault type;

[0010] A location module, configured to establish an associated graph of the strapping machine components according to the abnormal state identifier of the strapping machine and the determination result of the fault type to obtain a fault cause location result and a predicted value of the fault development trend;

[0011] A calculation module, configured to calculate the maintenance priority in combination with the strapping machine maintenance database according to the fault cause location result and the predicted value of the fault development trend to obtain a maintenance plan and an operation time suggestion;

[0012] A prediction module, configured to record the actual maintenance result and the predicted deviation according to the operation effect of the strapping machine after the implementation of the maintenance plan to obtain an updated fault feature library of the strapping machine and maintenance knowledge parameters.

[0013] In the third aspect of the present invention, a computer device is provided, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor invokes the instructions in the memory to cause the computer device to execute the above-mentioned belt machine fault monitoring method based on the Internet of Things.

[0014] In the fourth aspect of the present invention, a computer-readable storage medium is provided, in which instructions are stored. When it runs on a computer, it causes the computer to execute the above-mentioned belt machine fault monitoring method based on the Internet of Things.

[0015] In the technical solution provided by this application, multi-source operation data of each key part of the strapping machine is collected through a sensor network, the data is refined and processed, and a complete fault diagnosis and maintenance system is established, achieving remarkable technical effects. The acquisition mechanism of multi-source operation data solves the problem of the limitation of single-sensor information. The fusion of multi-dimensional data such as motor temperature data, tension control parameters, displacement data of the bundling manipulator, operating state of the cutting device, and strapping deformation rate provides a comprehensive operating state portrait of the strapping machine, greatly improving the integrity and accuracy of monitoring; the data is filtered and segmented to obtain the state characteristic data of the strapping machine with timestamps, effectively eliminating the interference of environmental noise and irrelevant signals, and enhancing the reliability of feature extraction; based on the state characteristic data of the strapping machine, the operating state is analyzed and compared, and an artificial intelligence model with adaptive thresholds and multi-feature fusion is adopted. When determining the fault type, machine learning algorithms are combined with expert knowledge, significantly reducing the probability of missed reports and false alarms; fourth, an associated graph of strapping machine components is established for fault cause location, and a causal reasoning algorithm is introduced. This algorithm fully considers the physical connection and functional dependence relationships between components, making fault location more accurate and capable of predicting the development trend of faults in advance, providing an adequate time window for maintenance decision-making; fifth, the maintenance priority is calculated in combination with the maintenance database to form a scientific maintenance plan and operation time suggestions. The multi-objective optimization algorithm balances the requirements of equipment safety, maintenance resources, and production plans, reducing unplanned downtime; most importantly, this invention establishes a complete knowledge feedback closed-loop. By recording the actual maintenance results and prediction deviations, the fault feature library and maintenance knowledge parameters are continuously updated. The self-learning ability of the system enables the accuracy of fault diagnosis to gradually improve with the usage time. In the specific functional field of the strapping machine, this invention gives full play to the advantages of artificial intelligence technologies such as adaptive filtering algorithms, causal propagation algorithms, and multi-objective optimization algorithms. Especially the application of the causal propagation algorithm in fault tracing breaks through the limitation that traditional correlation analysis is difficult to identify the root causes of hidden faults, enabling the system to handle complex fault propagation chains, providing a new technical idea for industrial equipment fault diagnosis, realizing the transformation from passive response to proactive prevention of the maintenance mode, significantly extending the effective operation time of the strapping machine, and improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a schematic diagram of an embodiment of the method for monitoring the faults of a strapping machine based on the Internet of Things in the embodiments of this application;

[0018] Figure 2 This is a schematic diagram of an embodiment of the belt tying machine fault monitoring system based on the Internet of Things in the embodiments of the present application;

[0019] Figure 3 This is a structural schematic block diagram of a computer device in the embodiments of the present invention. Detailed implementation manners

[0020] The embodiments of the present application provide a belt tying machine fault monitoring method and system based on the Internet of Things. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the term "comprising" or "having" and any variation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 , an embodiment of the belt tying machine fault monitoring method based on the Internet of Things in the embodiments of the present application includes:

[0022] Step S101: Collect operation data of each key part of the belt tying machine through a sensor network to obtain multi-source operation data including motor temperature data, tension control parameters, displacement data of the bundling robotic arm, operating status of the cutting device, and belt deformation rate;

[0023] Step S102: Filter and segment the data according to the multi-source operation data to obtain belt tying machine state characteristic data with time stamps;

[0024] Step S103: Analyze and compare the operating status of the belt tying machine based on the belt tying machine state characteristic data to obtain an abnormal state identifier of the belt tying machine and a fault type determination result;

[0025] Step S104: Establish a component association map of the belt tying machine according to the abnormal state identifier of the belt tying machine and the fault type determination result to obtain a fault cause location result and a fault development trend prediction value;

[0026] Step S105: Calculate the maintenance priority in combination with the belt tying machine maintenance database based on the fault cause location result and the fault development trend prediction value to obtain a maintenance plan and an operation time suggestion;

[0027] Step S106: According to the running effect of the strapping machine after the implementation of the maintenance plan, record the actual maintenance results and prediction deviations to obtain an updated fault feature library and maintenance knowledge parameters of the strapping machine.

[0028] It can be understood that the execution subject of this application can be an Internet of Things-based strapping machine fault monitoring system, or a terminal or a server, and specific limitations are not made here. In this embodiment of the application, the server is taken as an example of the execution subject for illustration.

[0029] Specifically, the operation data of each key part of the strapping machine collected by the sensor network is the basis of the whole method. Specifically, a temperature sensor and a current sensor are installed on the motor drive device of the strapping machine to collect the surface temperature of the motor and the current waveform data; a tension sensor and a displacement sensor are arranged on the tension control mechanism to record the change of the strapping tension and the displacement of the control rod; an acceleration sensor and an angle sensor are configured on the bundling robotic arm to measure the motion state and working angle of the robotic arm; a vibration sensor and a sound sensor are installed on the cutting device to monitor the vibration frequency and noise characteristics during the cutting process; the deformation of the strapping during the bundling process is measured by an optical sensor. These data are transmitted to the data processing unit through an industrial-grade wireless communication network to form multi-source operation data. For example, the motor temperature sensor collects data once per second, records the surface temperature of the motor as 85°C, and at the same time the current sensor records that the current waveform shows periodic fluctuations; the tension sensor detects that the strapping tension is 450 N, and the displacement sensor shows that the displacement of the control rod is 2.8 mm; the acceleration sensor measures that the acceleration of the robotic arm is 0.5 m / s², and the angle sensor measures that the working angle is 75°; the vibration sensor detects that the vibration frequency of the cutting device is 120 Hz, and the sound sensor records that the noise decibel value is 78 dB; the optical sensor measures that the deformation rate of the strapping is 0.15.

[0030] Filtering and segmenting multi-source operation data is to extract effective information. The noise and interference signals in the data are removed through an adaptive filtering algorithm to obtain a pure data stream. The adaptive filtering algorithm dynamically adjusts the filtering parameters according to the signal characteristics, such as smoothing the sudden interference in the motor temperature data. Then, the pure data stream is divided into data segments in the form of a sliding window with a width of 5 seconds and an overlap rate of 50% to form segmented data. An accurate time mark is added to each data segment to construct a time-aligned data frame. Time-domain feature quantities are extracted from the time-aligned data frame, including the mean value, peak value, root mean square value, kurtosis, and skewness. At the same time, the data frame is subjected to frequency-domain conversion to obtain the spectral energy distribution and the main frequency component information. The time-domain feature quantities and the frequency-domain information are feature-fused to generate the belt machine state feature data with timestamps. For example, after processing the motor temperature data, the average temperature in this period is 83.5°C, the peak value is 87.2°C, the root mean square value is 84.1°C, the kurtosis of the temperature data is 2.3, the skewness is 0.5, and the frequency-domain analysis shows that the main frequency components are concentrated in the low-frequency band. Analyzing and comparing the operating state of the belt machine is to detect abnormal conditions in a timely manner. A reference parameter library for the normal operation of the belt machine is constructed from historical data, including the standard feature distribution boundary values under various working conditions. Then, the difference between the belt machine state feature data and the corresponding data in the reference parameter library is calculated to obtain a feature deviation matrix. An abnormal weight coefficient is set according to the deviation degree of each parameter in the feature deviation matrix to generate a comprehensive abnormal index. By comparing the threshold, it is judged whether the comprehensive abnormal index exceeds the preset warning threshold to form an abnormal state identifier for the belt machine. The feature deviation matrix is matched for similarity according to the typical fault feature template of the belt machine to identify the target fault category. For example, when it is detected that the average motor temperature exceeds the upper limit of the normal range by 5°C, the high-frequency components in the current waveform increase, and the vibration sensor data shows abnormal vibration frequencies, the system calculates a comprehensive abnormal index of 0.78, which exceeds the warning threshold of 0.65, and identifies the motor bearing fault type through similarity matching.

[0031] The establishment of the component association map of the strapping machine is to accurately locate the cause of the fault. The component association map containing the physical connections and functional dependencies between components is constructed using the structural knowledge of the strapping machine. Abnormal component nodes are marked in the map according to the fault type determination result. The reverse tracing of the abnormal signal transmission path is carried out in the map through the causal propagation algorithm, and the possibility values of each relevant component as the root cause of the fault are calculated. The target fault component and its influence range are determined by sorting, and the fault cause location result is generated. Based on the development law in historical similar fault cases, the time series extrapolation calculation is carried out on the fault location result, and the fault severity change curve is constructed. Combining the fault severity change curve with the equipment safety threshold, the fault development time window and the potential risk level are predicted. For example, when a motor bearing fault is identified, the system marks the motor bearing node as abnormal in the association map. Through the causal propagation algorithm calculation, it is found that the connected transmission shaft and pulley are also affected, but the root cause of the fault is located at the bearing, and the possibility value is 0.85, which is much higher than other components. Historical data shows that if this type of fault is not handled, the motor will be completely damaged within 120 hours, and the predicted risk level is high.

[0032] Combining the maintenance database of the strapping machine to calculate the maintenance priority is to scientifically arrange the maintenance work. Historical maintenance cases matching the fault cause location result are extracted from the maintenance database to construct a maintenance knowledge base set. The fault development trend is predicted according to the maintenance knowledge base set to obtain the fault development trend prediction value, and the fault risk index is calculated according to the fault development trend prediction value. The fault risk index is weighted and combined with the production task priority, spare parts inventory status and technician availability to generate the maintenance urgency index. The maintenance tasks are classified through the maintenance urgency index to determine the maintenance operation level. The corresponding standard maintenance process is extracted according to the maintenance knowledge base set and the maintenance operation level to form a maintenance plan. Combining the fault development trend prediction value with the production plan arrangement, the optimal maintenance time window is determined, and the operation time suggestion is generated. For example, for the identified motor bearing fault, the system extracts similar cases from the maintenance database, calculates the fault risk index as 0.82. Considering that the current production plan urgency is medium, the spare parts inventory is sufficient, and the technicians are available on weekdays, the maintenance urgency index is generated as 0.75, which is determined as a secondary maintenance operation. It is recommended to replace the bearing within 48 hours and carry out the maintenance during the production idle period of the next weekend.

[0033] Recording the actual maintenance results and the prediction deviation is for continuously optimizing the monitoring system. Performance index data of the belt conveyor after the implementation of the maintenance plan is collected through the sensor network, and an evaluation report on the operation effect after maintenance is constructed. The evaluation report on the operation effect after maintenance is compared and analyzed with the state before maintenance, and the improvement metric value of the maintenance effect is calculated. The fault cause location result is compared with the actual fault source found in the maintenance, and the fault diagnosis accuracy evaluation data is recorded. According to the deviation situation in the fault diagnosis accuracy evaluation data, the feature threshold of the corresponding fault mode in the fault feature library of the belt conveyor is adjusted. In response to the difference between the predicted value of the fault development trend and the actual fault development trajectory, the time series parameters in the prediction algorithm are updated. The associated data of the maintenance effect improvement metric value and the maintenance plan are integrated and written into the maintenance knowledge parameter library to form an updated fault feature library and maintenance knowledge parameters. For example, after replacing the motor bearing, the temperature sensor shows that the motor temperature has dropped back to the normal range, the vibration frequency has returned to normal, and the comprehensive performance has increased by 90%. The actual maintenance finds that the bearing is indeed the root cause of the fault, and the diagnosis accuracy is 95%. However, the fault development speed is 15% slower than the prediction. Therefore, the system correspondingly adjusts the feature threshold of the motor bearing fault in the fault feature library and the time series parameters in the prediction algorithm to improve the accuracy of future diagnosis.

[0034] In the embodiments of the present application, multi-source operation data of each key part of the strapping machine is collected through a sensor network, the data is refined, and a complete fault diagnosis and maintenance system is established, achieving remarkable technical effects. The acquisition mechanism of multi-source operation data solves the problem of the limitation of single-sensor information. The fusion of multi-dimensional data such as motor temperature data, tension control parameters, displacement data of the bundling manipulator, operating state of the cutting device, and strapping deformation rate provides a comprehensive operation state portrait of the strapping machine, greatly improving the integrity and accuracy of monitoring; the data is filtered and segmented to obtain the state characteristic data of the strapping machine with time stamps, effectively eliminating the interference of environmental noise and irrelevant signals, and improving the reliability of feature extraction; based on the state characteristic data of the strapping machine, the operation state is analyzed and compared, and an artificial intelligence model with adaptive thresholds and multi-feature fusion is adopted. When determining the type of fault, machine learning algorithms are combined with expert knowledge, significantly reducing the probability of missed reports and false alarms; fourth, an associated component map of the strapping machine is established for fault cause location, and a causal reasoning algorithm is introduced. This algorithm fully considers the physical connection and functional dependence relationships between components, making fault location more accurate and capable of predicting the development trend of faults in advance, providing a sufficient time window for maintenance decisions; fifth, the maintenance priority is calculated in combination with the maintenance database to form a scientific maintenance plan and operation time suggestions. The multi-objective optimization algorithm balances the three aspects of equipment safety, maintenance resources, and production plans, reducing unplanned downtime; most importantly, the present invention establishes a complete knowledge feedback closed-loop. By recording the actual maintenance results and prediction deviations, the fault feature library and maintenance knowledge parameters are continuously updated. The self-learning ability of the system enables the accuracy of fault diagnosis to gradually improve with the use time. In the specific functional field of the strapping machine, the present invention gives full play to the advantages of artificial intelligence technologies such as adaptive filtering algorithms, causal propagation algorithms, and multi-objective optimization algorithms. Especially the application of the causal propagation algorithm in fault tracing breaks through the limitation that traditional correlation analysis is difficult to identify the root causes of hidden faults, enabling the system to handle complex fault propagation chains, providing a new technical idea for industrial equipment fault diagnosis, realizing the transformation of the maintenance mode from passive response to active prevention, significantly extending the effective operation time of the strapping machine, and improving production efficiency and product quality.

[0035] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0036] (1) Install temperature sensors and current sensors on the motor drive device of the strapping machine to collect the surface temperature and current waveform of the motor, and obtain motor temperature data;

[0037] (2) Arrange tension sensors and displacement sensors on the tension control mechanism of the strapping machine to record the tension change and control rod displacement of the strapping machine, and form tension control parameters;

[0038] (3)Configure an acceleration sensor and an angle sensor on the bundling robotic arm of the strapping machine to measure the motion state and working angle of the robotic arm and generate displacement data of the bundling robotic arm;

[0039] (4)Install a vibration sensor and a sound sensor on the cutting device of the strapping machine to monitor the vibration frequency and noise characteristics during the cutting process and generate the operating state of the cutting device;

[0040] (5)Measure the deformation of the strap during the bundling process through an optical sensor, calculate the degree of force-induced deformation of the strap, and obtain the strap deformation rate;

[0041] (6)Transmit the motor temperature data, tension control parameters, displacement data of the bundling robotic arm, operating state of the cutting device, and strap deformation rate to the data processing unit through an industrial-grade wireless communication network to form multi-source operation data with corresponding time stamps.

[0042] Specifically, install a temperature sensor and a current sensor on the motor drive device of the strapping machine to collect the surface temperature of the motor and the current waveform to obtain the motor temperature data. The motor drive device is the core power source of the strapping machine and is responsible for providing the power required for the operation of the strapping machine. The temperature sensor uses the thermocouple or thermistor principle and is directly attached to the surface of the motor to regularly sample the surface temperature of the motor at a sampling frequency of 1 time per second; the current sensor is based on the Hall effect principle and is installed around the motor power line to record the current change waveform during the operation of the motor in real time at a sampling frequency of 100 times per second. Through the combined use of these two sensors, a set of motor temperature data containing temperature scalars and current vectors is formed, effectively reflecting the working state of the motor. Arrange a tension sensor and a displacement sensor on the tension control mechanism of the strapping machine to record the tension change of the strapping machine and the displacement of the control rod to form the tension control parameters. The tension control mechanism is a key component in the strapping machine that adjusts the tightening force of the strap and has a direct impact on the quality of the strap. The tension sensor uses the strain gauge principle and is fixed on the bearing seat of the tension adjusting wheel to measure the force on the tension wheel, and then calculate the strap tension value at a sampling frequency of 20 times per second; the displacement sensor is based on the linear variable differential transformer principle and is installed on the movable part of the control rod to accurately measure the displacement change of the control rod during operation at a sampling frequency of 50 times per second. After these two sets of data are time-aligned, a set of parameters reflecting the tension control state of the strapping machine is formed.

[0043] The bundling robotic arm of the strapping machine is equipped with an acceleration sensor and an angle sensor to measure the motion state and working angle of the robotic arm and generate displacement data of the bundling robotic arm. The bundling robotic arm is the actuator that realizes the strapping and fixing of the strap, and its motion accuracy and stability directly affect the bundling effect. The acceleration sensor is based on the piezoelectric effect principle and is installed at the robotic arm joint to simultaneously collect the motion acceleration of the robotic arm in three axes, with a sampling frequency of 200 times per second; the angle sensor uses the photoelectric encoder principle and is installed on the robotic arm rotating shaft to continuously monitor the rotation angle of the robotic arm, with an accuracy of 0.1 degrees and a sampling frequency of 100 times per second. Through the integral operation of the acceleration data and in combination with the angle data, a complete three-dimensional motion trajectory model of the robotic arm is constructed.

[0044] A vibration sensor and a sound sensor are installed in the cutting device of the strapping machine to monitor the vibration frequency and noise characteristics during the cutting process and generate operation state data of the cutting device. The cutting device is responsible for performing the cutting operation after the strap is bundled, and its working state affects the quality of the strap cut. The vibration sensor is based on the piezoelectric effect principle and is directly fixed on the base of the cutting mechanism to collect the vibration signal during the cutting process, with a frequency response range of 0 - 1000 Hz and a sampling frequency of 1000 times per second; the sound sensor uses the capacitive microphone principle and is installed near the cutting area to record the sound characteristics generated during the cutting process, with a frequency response range of 20 Hz - 20 kHz and a sampling frequency of 8000 times per second. After the vibration data and the sound data are subjected to spectral analysis, a feature vector reflecting the working state of the cutting device is formed.

[0045] The deformation of the strap during the bundling process is measured by an optical sensor, the degree of force-induced deformation of the strap is calculated, and the strap deformation rate is obtained. The optical sensor uses the laser ranging principle and is installed on a fixed bracket to vertically irradiate the surface of the strap and measure the thickness change of the strap before and after being stressed, with a sampling frequency of 50 times per second. The formula for calculating the strap deformation rate is as follows:

[0046]

[0047] where, represents the strap deformation rate, represents the initial thickness of the strap (mm), represents the thickness of the strap in the stressed state (mm), represents the strap material compensation coefficient (dimensionless), represents the strap elastic characteristic parameter, Represents the tension acting on the strap. Convert the thickness change data into a deformation rate indicator that reflects the material properties and actual usage status of the strap. Transmit the motor temperature data, tension control parameters, bundling robot arm displacement data, cutting device operating status and strap deformation rate to the data processing unit through the industrial wireless communication network to form multi-source operating data with corresponding time stamps. The industrial wireless communication network adopts industrial Internet of Things protocols, such as MQTT or OPC UA, to ensure the reliability and real-time performance of data transmission. After receiving the scattered sensor data, the data processing unit performs time synchronization processing and adds a unified timestamp mark to each set of data with a time accuracy of milliseconds; then the data format is unified and different types of sensor data are converted into a standard format; a multidimensional data matrix is ​​created, in which each row represents the complete state at a time point and each column represents the specific parameters of different sensors.

[0048] For example: after the strapping machine is started, the motor temperature gradually rises from the normal temperature of 25℃, and stabilizes at 78℃ after 10 minutes. At the same time, the current sensor records that the motor working current fluctuates between 4.8A-5.2A, and the waveform shows typical periodic characteristics; the tension sensor measures that the strap tension is stable at 420N, and the displacement sensor records that the control rod displacement gradually increases from 0mm to 2.5mm during the adjustment process and then remains stable; the acceleration sensor measures that the acceleration of the mechanical arm reaches 0.6m / s² at the beginning of the bundling action, and the angle sensor records that the mechanical arm rotates from the initial position 0° to the working position 85°; the vibration sensor detects a vibration signal with a peak frequency of 150Hz at the moment of cutting, and the sound sensor records 75dB of noise; the optical sensor measures that the initial thickness of the strap is 0.8mm, which is compressed to 0.68mm under the tension of 420N. Substituting it into the formula, the deformation rate of the strap is 0.15, which is within the normal range. These data are transmitted to the data processing unit in real time through the industrial wireless network, forming a complete multi-source operation data matrix, providing a basis for subsequent fault analysis. In the data matrix, a complete status record is generated every 100ms, including all sensor readings at the current moment, and is accompanied by a precise timestamp, such as "2024-10-15 14:30:25.700", which facilitates time series comparison and correlation analysis in subsequent analysis.

[0049] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0050] (1) Use adaptive filtering algorithms to remove noise and interference signals from multi-source operating data to obtain pure data streams;

[0051] (2) Divide the pure data stream into data segments using a sliding window method with a width of 5 seconds and an overlap rate of 50% to form segmented processing data;

[0052] (3) The time synchronization mechanism adds accurate time stamps to the segmented data to construct a time-aligned data frame;

[0053] (4) Extract time-domain feature quantities including mean value, peak value, root mean square value, kurtosis, and skewness from the time-aligned data frame through time-domain analysis methods;

[0054] (5) Perform frequency-domain conversion on the time-aligned data frame to obtain the spectral energy distribution and main frequency component information;

[0055] (6) Perform feature fusion on the time-domain feature quantities, spectral energy distribution, and main frequency component information to generate the state feature data of the strapping machine with time stamps.

[0056] Specifically, the noise and interference signals in the multi-source operation data are removed through an adaptive filtering algorithm to obtain a pure data stream. The adaptive filtering algorithm is a signal processing method that can dynamically adjust the filtering parameters according to the signal characteristics. The specific implementation uses the least mean square error (LMS) adaptive filter. This filter has a good suppression effect on the sudden interference in the motor temperature data of the strapping machine, the high-frequency jitter in the tension control parameters, the random fluctuations in the displacement data of the bundling manipulator, the environmental noise in the vibration signal of the cutting device, and the optical interference in the measurement of the strapping deformation rate. The filtering process estimates the noise characteristics in the signal, and then automatically adjusts the filter coefficients according to the characteristic differences between the noise and the useful signal to achieve precise removal of the noise. For different types of sensor data, the filter parameters will be automatically adjusted. For example, for signals with slow changes such as motor temperature, a narrower passband is used; for signals with rapid changes such as manipulator displacement, a wider passband is used to ensure that the filtered data removes the interference and retains the effective information of the original signal. The pure data stream is divided into data segments in the form of a sliding window with a width of 5 seconds and an overlap rate of 50%. The sliding window is a technique that divides continuous time series data into multiple overlapping segments. The window width of 5 seconds means that each data segment contains 5 seconds of continuous data, and the 50% overlap rate means that adjacent two windows share half of the data points. Selecting 5 seconds as the window width is based on the working cycle characteristics of the strapping machine. Generally, the single strapping operation cycle is between 3 - 8 seconds, and the 5-second window can completely capture most of the operation characteristics; while the 50% overlap rate ensures the continuity of the data and avoids the problem of information loss at the window boundaries. In specific implementation, if the sensor sampling frequency is 100Hz, each 5-second window contains 500 data points, and adjacent windows share 250 data points. This division method not only ensures the locality of data analysis but also maintains the continuity of the signal, which is suitable for capturing the state change characteristics during the working process of the strapping machine.

[0057] The time synchronization mechanism adds accurate time stamps to the segmented data to construct time-aligned data frames. The time synchronization mechanism solves the problem of inconsistency in the time dimension of multi-source heterogeneous sensor data. The Network Time Protocol (NTP) is used to ensure the clock synchronization of all sensor nodes, with the error controlled within the millisecond level; then, time stamps are added to the segmented data, recording the start time and end time of each window; through time interpolation and resampling techniques, sensor data with different sampling rates are aligned to a unified time scale. For example, the sampling rate of the motor temperature sensor is 1 Hz, while the sampling rate of the vibration sensor is 1000 Hz. The time synchronization mechanism will perform linear interpolation on the motor temperature data to create temperature estimates consistent with the time points of the vibration data, ensuring the precise correspondence of different types of data in the data frame in terms of time. Time-domain feature quantities including mean, peak, root mean square value, kurtosis, and skewness are extracted from the time-aligned data frame through time-domain analysis methods. Time-domain feature quantities are statistical indicators reflecting the time distribution characteristics of signals and are of great significance for the fault diagnosis of the banding machine. The mean calculates the average value of all sampling points within the data segment, reflecting the central tendency of the signal; the peak extracts the maximum absolute value in the data segment, reflecting the extreme value characteristics of the signal; the root mean square value calculates the square root of the mean of the sum of the squares of the data points, reflecting the energy level of the signal; kurtosis measures the sharpness of the data distribution, and the kurtosis value under normal working conditions is usually close to 3 and will deviate significantly under abnormal conditions; skewness measures the asymmetry of the data distribution, reflecting the inclination characteristics of the signal waveform. These time-domain feature quantities together constitute a feature vector describing the statistical characteristics of the data segment, which can effectively distinguish different working states and fault modes of the banding machine.

[0058] Perform a frequency-domain transformation on the time-aligned data frame to obtain the spectral energy distribution and the information of the main frequency components. Frequency-domain analysis converts the time-domain signal into a frequency-domain representation through the Fast Fourier Transform (FFT), revealing the frequency composition of the signal. For the banding machine, different components will exhibit different spectral characteristics under normal working and faulty conditions. For example, a motor bearing fault will generate abnormal resonance peaks at specific frequencies. The frequency-domain transformation applies a Hanning window function to the data segment to reduce spectral leakage, and then performs the FFT operation to obtain the complex spectrum, and calculates the Power Spectral Density (PSD) to obtain the distribution of signal energy at each frequency. Extract features such as the total energy, the energy ratio of each frequency band, the main frequency point and its amplitude from the PSD. These features are particularly important for identifying the periodic faults and harmonic characteristics of the banding machine. Fuse the time-domain feature quantities with the spectral energy distribution and the information of the main frequency components to generate the state feature data of the banding machine with time stamps. Feature fusion is a process of integrating features of different dimensions and scales into a unified representation. Adopting a feature-level fusion strategy, the time-domain feature vector and the frequency-domain feature vector are concatenated into a high-dimensional feature vector, and at the same time, feature normalization is performed to ensure the balanced weights of features with different dimensions during the fusion process. The fused feature data retains the time stamp information of the original window, facilitating subsequent time series analysis and fault tracking. The generated state feature data of the banding machine is a multi-dimensional time series data set, and each record contains a complete state feature description of the banding machine within a time window.

[0059] For example, when the banding machine is performing a bundling task, the vibration sensor detects abnormal vibrations generated by the cutting device. The original vibration signal is mixed with environmental noise. After being processed by the adaptive filtering algorithm, the 50Hz power supply interference and random white noise are successfully filtered out, and the vibration characteristics of the device itself are retained. The filtered 5-second data is divided according to a sliding window, and a new data segment is formed every 2.5 seconds, and adjacent data segments share half of the data. Add accurate time stamps to each data segment, such as "2024-10-15 14:30:25.500-14:30:30.500". Through time-domain analysis, it is calculated that the mean value of the vibration signal is close to 0, the peak value is 2.8g, the root mean square value is 0.9g, the kurtosis value is 4.2 (higher than the normal value of 3, indicating the presence of impact components), and the skewness is 0.3 (slightly right-skewed). Perform FFT analysis on this data segment and find that there is an obvious frequency peak at 150Hz, and the energy is 3 times that of the fundamental frequency, which is significantly different from the spectral characteristics under normal conditions. After fusing these time-domain and frequency-domain features, a state feature vector containing 25 feature quantities is formed, along with the accurate time stamp of this time window. By comparing with historical data, this feature combination highly matches the fault mode of cutting tool wear.

[0060] In a specific embodiment, the process of performing step S103 may specifically include the following steps:

[0061] (1) Construct a benchmark parameter library for the normal operation of the belt tying machine from historical data, including the standard characteristic distribution boundary values under various working conditions;

[0062] (2) Calculate the differences between the state characteristic data of the belt tying machine and the corresponding data in the benchmark parameter library for the normal operation of the belt tying machine to obtain a characteristic deviation matrix;

[0063] (3) Set abnormal weight coefficients based on the deviation degrees of the parameters in the characteristic deviation matrix to generate a comprehensive abnormal index;

[0064] (4) Determine whether the comprehensive abnormal index exceeds the preset warning threshold through a threshold comparison method to form an abnormal state identifier for the belt tying machine;

[0065] (5) Perform similarity matching on the characteristic deviation matrix according to the typical fault characteristic template of the belt tying machine to identify the target fault category;

[0066] (6) Combine the abnormal state identifier of the belt tying machine with the target fault category to form the abnormal state identifier and fault type determination result of the belt tying machine.

[0067] Specifically, a normal operation benchmark parameter library for the strapping machine is constructed from historical data, which includes the standard feature distribution boundary values under various working conditions. The normal operation benchmark parameter library is a set of characteristic data of the normal operation state of the strapping machine under different working conditions, and is constructed by collecting and statistically analyzing the data of the strapping machine running stably for a long time. The specific construction process includes: collecting the data of the strapping machine that has run without faults for at least 30 consecutive days; classifying the data according to different working load conditions (light load, medium load, heavy load), different types of bundling materials (plastic straps, metal straps), and different running speeds (low speed, medium speed, high speed), etc.; calculating the statistical distribution characteristics of each characteristic parameter (such as motor temperature, vibration spectrum, tension parameter, etc.) under each type of working condition, including mean value, standard deviation, maximum value, and minimum value; determining the normal fluctuation range of each characteristic parameter according to the 3σ principle to form the standard feature distribution boundary values. The obtained benchmark parameter library is a multi-dimensional lookup table, which can quickly locate the corresponding normal parameter range according to the current working condition. Calculate the difference between the state characteristic data of the strapping machine and the corresponding data in the normal operation benchmark parameter library of the strapping machine to obtain the characteristic deviation matrix. The characteristic deviation matrix is a quantitative representation of the degree to which the current running state of the strapping machine deviates from the normal state. The difference calculation retrieves the corresponding standard feature distribution boundary values from the benchmark parameter library according to the working condition of the current strapping machine (such as the type of strap being processed, the current load, etc.); then calculates the normalized deviation between the current characteristic data and the standard value, that is, the actual value of each characteristic parameter minus the corresponding standard mean value, and then divides by the standard deviation to obtain the standardized deviation value; organize the standardized deviation values of all characteristic parameters into a matrix form, with rows representing different time windows and columns representing different characteristic parameters. The characteristic deviation matrix intuitively reflects the degree to which each parameter deviates from the normal state at each time point, providing a basis for anomaly detection.

[0068] Set the anomaly weight coefficient based on the deviation degree of each parameter in the characteristic deviation matrix to generate the comprehensive anomaly index. Different characteristic parameters have different contributions to fault identification, so different weights need to be assigned. The anomaly weight coefficient is set based on expert experience to specify the basic weight for each characteristic parameter, reflecting its importance for fault indication; then introduce an adaptive mechanism to dynamically adjust the weight according to the sensitivity of each parameter in the historical fault data; combine the non-linear mapping of the current deviation degree to assign a higher weight to the parameters with severe deviation. The comprehensive anomaly index is the weighted sum result after multiplying each element in the characteristic deviation matrix by the corresponding anomaly weight coefficient, and is a scalar between 0 and 1. The closer the value is to 1, the more severe the anomaly degree is.

[0069] Determine whether the comprehensive anomaly index exceeds the preset warning threshold to form an anomaly status identifier for the strapping machine. Threshold comparison is a binary classification decision method that determines whether the current status is normal or abnormal based on the magnitude relationship between the comprehensive anomaly index and the preset threshold. The setting of the warning threshold adopts a hierarchical strategy, including three levels: attention level (0.6), warning level (0.75), and danger level (0.85). During threshold comparison, the calculated comprehensive anomaly index is compared with these preset thresholds one by one to determine the warning level of the current status; if the comprehensive anomaly index is lower than the lowest threshold, it is determined to be in a normal state; if it exceeds a certain level threshold, an anomaly status identifier of the corresponding level is formed, including information such as the anomaly level, occurrence time, and related parameters.

[0070] Perform similarity matching on the feature deviation matrix according to the typical fault feature template of the strapping machine to identify the target fault category. The fault feature template is a set of feature deviation patterns for various known faults, obtained through learning historical fault cases. The similarity matching uses the cosine similarity calculation method to calculate the similarity degree between the current feature deviation matrix and each fault template. The formula is as follows:

[0071]

[0072] where, represents the fault similarity, represents the deviation value of the j-th feature in the i-th time window of the feature deviation matrix, represents the corresponding feature deviation value in the k-th fault template, represents the number of time windows, represents the number of feature parameters, represents the weight coefficient of the p-th key feature, represents the flag indicating whether the feature j is a key indicator feature in the fault type k (1 means yes, 0 means no), represents the total number of key features. It not only considers the similarity of the overall deviation pattern but also particularly emphasizes the key indicator features of the fault type, improving the accuracy of identification. After the similarity calculation is completed, the fault type with the highest similarity and exceeding the threshold of 0.8 is selected as the target fault category. If all similarities are lower than the threshold, it is marked as "unknown fault type". Combine the anomaly status identifier of the strapping machine with the target fault category to form the anomaly status identifier of the strapping machine and the fault type determination result. This is the process of integrating the results of two subtasks, anomaly detection and fault classification, into the output. The combination method uses a structured data format, including the following fields: anomaly status identifier (normal / attention / warning / danger), anomaly occurrence time, anomaly duration, target fault category, similarity score, main anomaly parameters involved and their deviation values, recommended processing priority.

[0073] For example: During the operation of a certain bundling machine, the data processing unit received preprocessed state characteristic data. Comparing with the normal operation benchmark parameter library, it was found that the current working condition was "medium load, plastic bundling, medium speed", and the standard parameter boundary values under this working condition were extracted from the library. Then, the difference between the current characteristics and the standard values was calculated to obtain the characteristic deviation matrix. Among them, the motor temperature deviation was +2.3σ (indicating 2.3 standard deviations beyond the normal average), the energy deviation of the cutting device vibration spectrum at 150 Hz was +4.1σ, the noise characteristic deviation was +2.8σ, while the deviation of the tension control parameter and the bundling deformation rate were both within 1σ. According to the preset weight coefficients (vibration spectrum 0.35, noise characteristic 0.25, motor temperature 0.2, tension parameter 0.1, deformation rate 0.1), the calculated comprehensive anomaly index was 0.82, exceeding the warning level threshold of 0.75. Then, the similarity matching calculation was performed between the characteristic deviation matrix and each template in the fault characteristic template library, and the similarity with the "cutting tool wear" template was 0.91, much higher than other fault types (such as "motor bearing fault" 0.43, "tension control disorder" 0.38, etc.). The formed determination result was: warning level abnormal state, occurrence time 2024-10-15 14:35:22, duration 8 minutes, target fault category cutting tool wear, similarity 0.91, main abnormal parameters 150 Hz vibration spectrum (+4.1σ) and noise characteristic (+2.8σ), recommended processing priority medium, and it was recommended to arrange maintenance within 24 hours.

[0074] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0075] (1) Construct a bundling machine component association graph containing the physical connections and functional dependencies between components by using the bundling machine structure knowledge;

[0076] (2) Mark the abnormal component nodes in the bundling machine component association graph according to the fault type determination result;

[0077] (3) Retroactively trace the abnormal signal transmission path in the bundling machine component association graph through the causal propagation algorithm, and calculate the possibility values of each relevant component as the root cause of the fault;

[0078] (4) Determine the target fault component and its influence range from the sorted possibility values, and generate the fault cause location result;

[0079] (5) Based on the development law in historical similar fault cases, perform time series extrapolation calculation on the fault cause location result, and construct a fault severity change curve;

[0080] (6) Combine the fault severity change curve with the equipment safety threshold to predict the fault development time window and potential risk level, and form a fault development trend prediction value.

[0081] Specifically, the knowledge of the strapping machine structure is used to construct a component association graph of the strapping machine that includes the physical connections and functional dependencies between components. The component association graph is a network model that describes the internal structure and functional relationships of the strapping machine, represented by a directed graph, where the nodes represent the various physical components of the strapping machine (such as motors, drive shafts, tension controllers, cutting devices, etc.), and the edges represent the relationships between components. The relationship types include physical connection relationships (such as component A is directly connected to component B) and functional dependency relationships (such as the function of component C depends on the normal operation of component D). The graph construction process includes: analyzing the structure drawings of the strapping machine, identifying all key components and setting them as nodes; determining the physical connection relationships between components and establishing connection edges; analyzing the working mechanisms between components, determining the functional dependency relationships, and establishing dependency edges; assigning a transmission coefficient to each edge to represent the attenuation degree of the fault signal transmitted from one component to another. The formed component association graph contains approximately 30 nodes and 60 directed edges, which completely describes the internal structure topology and functional hierarchy of the strapping machine. Mark the abnormal component nodes in the component association graph of the strapping machine according to the fault type determination result. The process of mapping the fault type determination result onto the component association graph. According to the correspondence table between the fault type and the components, determine the components directly related to the current fault type, and mark these component nodes as abnormal nodes in the graph. When marking, assign an initial abnormal intensity value to each abnormal node, which is determined based on the similarity score of the fault type determination and the correlation degree of the component with the fault type. For example, for the "cutting tool wear" fault type, mark the cutting tool node as abnormal and assign an initial abnormal intensity value of 0.91 according to the previously obtained similarity of 0.91.

[0082] Reverse trace the abnormal signal transmission path in the component association graph of the strapping machine through the causal propagation algorithm, and calculate the possibility values of each relevant component as the root cause of the fault. The causal propagation algorithm is a calculation method for analyzing signal transmission and influence scope in a network structure. In this solution, an improved belief propagation algorithm is adopted. The algorithm starts from the marked abnormal nodes and propagates backward along the edges in the graph, calculating the possibility of the upstream nodes associated with the abnormal nodes as the root cause of the fault; then consider the converging influence of multiple paths. For a component with multiple downstream abnormal nodes, its fault possibility is higher; through multiple rounds of iterative calculation, stop when the change in the possibility values of each node is less than the threshold or the maximum number of iterations is reached. Specifically in the calculation, the fault possibility of each node is affected by the weighted influence of all its downstream abnormal nodes, and the weight is determined by the transmission coefficient of the connection edge and the path length.

[0083] Determine the target faulty component and its influence scope based on the sorting of possibility values, and generate the result of fault cause localization. In this step, the fault possibility values of each component calculated by the causal propagation algorithm are sorted in descending order to identify the most likely root cause of the fault. According to the sorting result, select the component with the highest possibility value as the target faulty component; at the same time, determine the influence scope of this component based on the graph structure, including all downstream components that may be affected by this fault. The generated result of fault cause localization includes the information of the target faulty component, the fault possibility value, the list of influence scope, and the description of the causal chain between the fault and the observed anomaly. Based on the development law in historical similar fault cases, conduct a time series extrapolation calculation on the result of fault cause localization to construct a curve of fault severity change. In this step, use the similar fault cases recorded in the historical fault database to analyze the time law of fault development. Retrieve the historical cases in the database that match the current target faulty component and fault type; then extract the data points of the change of fault severity over time in these cases; then use the non-linear regression method to fit the mathematical model of fault development, and the commonly used models include the exponential growth model, the Weibull model or the polynomial model; substitute the severity value of the current fault into the model for time series extrapolation calculation to predict the change trend of fault severity in a future period of time, and form a curve of fault severity change. This curve takes time as the horizontal axis and fault severity as the vertical axis, intuitively showing the development speed and trend of the fault.

[0084] Combine the curve of fault severity change with the equipment safety threshold to predict the fault development time window and the potential risk level, and form the predicted value of fault development trend. The equipment safety threshold refers to the safe operation limit of each component of the banding machine at different severities. Exceeding this limit will lead to a decline in equipment performance, functional failure or safety accidents. In the prediction process, mark the safety threshold lines of different levels on the curve of fault severity change; then calculate the intersection points of the curve and each threshold line to obtain the predicted time when the fault develops to each severity level; then comprehensively evaluate the potential risk level according to the time urgency and consequence severity; form a predicted value of fault development trend including the fault development time window, the potential risk level, the predicted impact and the recommended handling time limit.

[0085] For example, a certain banding machine is detected with a "cutting tool wear" fault, and the similarity is 0.91. In the component association graph, the cutting tool node is marked as abnormal, and the initial abnormal intensity is 0.91. Through backward tracing using the causal propagation algorithm, the fault probability values of all relevant components are calculated: the cutting tool itself is 0.91, the tool drive motor is 0.75, the control circuit is 0.45, the transmission shaft is 0.68, and the connecting bolt is 0.72. After sorting, it is determined that the cutting tool is the most likely fault source, and its influence range includes cutting quality, cutting noise, and the treatment effect of the band end. 15 similar cases of tool wear are retrieved from the historical database, and the variation law of the fault severity over time in these cases is analyzed, and it is found that it conforms to the exponential growth model. Based on the initial severity of the current fault and the historical development law, the fault severity change curve within the next 7 days is predicted. The equipment safety thresholds are divided into three levels: attention level (0.6), warning level (0.8), and danger level (0.95). By comparing the safety threshold with the change curve, it is predicted that the fault will reach the warning level after 3 days and the danger level after 7 days. If not handled, it will lead to cutting failure and possible safety risks. The predicted value of the fault development trend formed is: faulty component - cutting tool, current severity - 0.5, expected to reach the warning level after 3 days, and the danger level after 7 days. It is recommended to arrange for tool replacement within 3 days. The potential risk level is medium, and the expected impact is a decrease in the quality of band cutting and possible cutting failure. This prediction result provides a clear time window and priority basis for maintenance decision-making.

[0086] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0087] (1) Extract historical maintenance cases that match the fault cause location result from the banding machine maintenance database to construct a maintenance knowledge base set;

[0088] (2) Predict the fault development trend based on the maintenance knowledge base set to obtain the predicted value of the fault development trend, and calculate the fault risk index according to the predicted value of the fault development trend;

[0089] (3) Perform a weighted combination of the fault risk index with the production task priority, spare part inventory status, and technician availability to generate a maintenance urgency index;

[0090] (4) Classify the maintenance tasks through the maintenance urgency index to determine the maintenance operation level;

[0091] (5) Extract the corresponding standard maintenance process based on the maintenance knowledge base set and the maintenance operation level to form a maintenance plan;

[0092] (6) Combine the predicted value of the fault development trend with the production plan arrangement to determine the best maintenance time window and generate an operation time suggestion.

[0093] Specifically, historical maintenance cases that match the fault cause location results are extracted from the belt machine maintenance database to construct a maintenance knowledge base set. The maintenance database is a structured data warehouse that records historical fault maintenance experiences and contains information such as fault types, maintenance methods, required resources, and maintenance effects. The extraction process uses case-based reasoning technology to retrieve similar historical cases based on the characteristic parameters of the current fault (fault components, fault types, fault severity, operating environment, etc.). The similarity calculation comprehensively considers the matching degree of fault components, the matching degree of fault types, the similarity of fault symptoms, and the similarity of operating environments, and uses a weighted summation method to obtain the comprehensive similarity. Historical cases with a similarity exceeding a preset threshold (usually 0.75) are selected to form the maintenance knowledge base set. Each case in the base set contains key information such as fault description, maintenance method, required tools and spare parts, maintenance duration, and maintenance effect evaluation. Based on the maintenance knowledge base set, the fault development trend is predicted to obtain a fault development trend prediction value, and the fault risk index is calculated according to the fault development trend prediction value. The fault development trend prediction is based on the time series data of fault development in historical cases, and a mathematical model is established through regression analysis to predict the development trajectory of the current fault. The prediction model uses a hybrid prediction method, and its calculation formula is as follows:

[0094]

[0095] Wherein, represents the fault risk index, represents the fault development rate in the i-th historical case, represents the current time, represents the time of the i-th historical case, represents the time decay coefficient, n represents the number of historical cases, represents the current fault severity, represents the maximum fault severity, represents the predicted time required for the fault to develop to the critical state, represents the time urgency coefficient, , , represent the weight coefficients of each factor and satisfy + + = 1. This formula comprehensively considers three factors: the weighted average of the historical case development rate (considering time decay), the current fault severity, and the time urgency of fault development, and generates a risk index value between 0 and 1. The larger the value, the higher the risk.

[0096] The failure risk index is combined with the production task priority, spare part inventory status, and technician availability in a weighted manner to generate a maintenance urgency index. The production task priority reflects the importance of the current strapping machine in the production plan, which is usually provided by the production management system and is divided into five levels (from 1 being the lowest to 5 being the highest); the spare part inventory status indicates the inventory situation of the parts required for maintenance, including information such as the inventory quantity and procurement cycle; the technician availability describes the current workload and schedulability of the personnel with the corresponding maintenance skills. The weighted combination uses a multi-factor evaluation model, considering the influence degree of each factor on the maintenance decision, to generate a comprehensive maintenance urgency index. This index is a value between 0 and 100, which intuitively reflects the urgency of the maintenance task. The maintenance tasks are classified according to the maintenance urgency index to determine the maintenance operation level. The maintenance operation level is a classification of the urgency and handling method of the maintenance task, generally divided into four levels: level 1 (emergency maintenance), level 2 (planned maintenance), level 3 (observational maintenance), and level 4 (routine maintenance). The classification process uses a threshold segmentation method to determine the corresponding maintenance operation level according to the numerical range of the maintenance urgency index. Generally speaking, an index value of 90 - 100 corresponds to level 1 maintenance, 70 - 89 corresponds to level 2 maintenance, 50 - 69 corresponds to level 3 maintenance, and 0 - 49 corresponds to level 4 maintenance. Different levels of maintenance operations have different response time requirements and resource allocation strategies. The corresponding standard maintenance process is extracted based on the maintenance knowledge base set and the maintenance operation level to form a maintenance plan. The standard maintenance process is an operating procedure preset for specific fault types and maintenance levels, including content such as repair steps, required tools and spare parts, and safety precautions. The extraction process selects the most matching repair cases for the current fault type from the maintenance knowledge base; then selects the corresponding processing flow template according to the maintenance operation level; then integrates the specific repair steps in the historical cases and the current maintenance requirements to form a customized maintenance plan. The generated maintenance plan is a structured document that details all the operation steps and resource requirements from preparation to completion and acceptance.

[0097] Combined with the predicted value of the fault development trend and the production plan arrangement, the optimal maintenance time window is determined to generate operation time suggestions. The optimal maintenance time window is the time period with the least impact on production under the premise of ensuring the safe operation of the equipment. The determination process calculates the time points when the fault develops to each dangerous level according to the predicted value of the fault development trend to determine the latest execution time of the maintenance; then obtains the production task arrangement of the recent strapping machine from the production plan management system to identify production gaps or low-load periods; comprehensively considering the time required for maintenance operations, the available time periods of technicians, and the arrival time of spare parts, selects the time period with the least impact on production within the safety time limit as the recommended maintenance time window. The generated operation time suggestions include the recommended start time, estimated completion time, production impact assessment, and scheduling suggestions.

[0098] For example: A certain belt tying machine is diagnosed with a fault of worn cutting tool, and the cause of the fault is determined to be severe wear of the cutting tool itself. Retrieve relevant cases from the maintenance database, and find 15 similar cases of worn cutting tools. Among them, 12 cases have a similarity exceeding 0.8, including key information such as maintenance methods, required spare parts, and average maintenance time, which constitute the basic set of maintenance knowledge. According to the data of these historical cases, the fault risk index is calculated to be 0.78 using the above formula. Currently, this belt tying machine is responsible for producing high-value products, and the production task priority is 4 (out of 5). The inventory of cutting tool spare parts required for maintenance is sufficient (10 pieces in stock, with a daily consumption of 0.5 pieces). The technical personnel with the corresponding maintenance skills have a moderate workload in the next three days. Considering these factors comprehensively, the maintenance urgency index is calculated to be 82, corresponding to the secondary maintenance operation level (planned maintenance). Extract the standard maintenance process applicable to cutting tool replacement from the basic set of maintenance knowledge, and form a maintenance plan including 6 main steps such as equipment shutdown, safety locking, removal of the old tool, installation of the new tool, adjustment of cutting parameters, and test verification, which is expected to take 2 hours to complete. According to the prediction of the fault development, the cutting tool will reach the warning level in 3 days and the dangerous level in 7 days. Querying the production plan reveals that there is a 4-hour production switching gap the day after tomorrow. Considering all factors comprehensively, the recommended operation time is: It is recommended to perform maintenance during the production switching gap (10:00 - 14:00) the day after tomorrow morning. The expected maintenance duration is 2 hours, and the impact on production is controllable. It is recommended to notify the production department in advance to make connection preparations.

[0099] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0100] (1) Collect the performance index data of the belt tying machine after the implementation of the maintenance plan through the sensor network, and construct an evaluation report on the operation effect after maintenance;

[0101] (2) Compare and analyze the evaluation report on the operation effect after maintenance with the state before maintenance, and calculate the improvement measurement value of the maintenance effect;

[0102] (3) Compare the fault cause location result with the actual fault source found in the maintenance, and record the evaluation data of the fault diagnosis accuracy;

[0103] (4) Adjust the feature threshold of the corresponding fault mode in the belt tying machine fault feature library according to the deviation situation in the evaluation data of the fault diagnosis accuracy;

[0104] (5) Update the time series parameters in the prediction algorithm for the difference between the predicted value of the fault development trend and the actual fault development trajectory;

[0105] (6) Integrate and write the associated data of the maintenance effect improvement measurement value and the maintenance plan into the maintenance knowledge parameter library to form an updated belt tying machine fault feature library and maintenance knowledge parameters.

[0106] Specifically, after the implementation of the maintenance plan, the performance index data of the strapping machine is collected through the sensor network to construct an evaluation report on the operation effect after maintenance. The same sensor network is used for data collection after maintenance as in the initial monitoring process, including motor temperature sensors, tension sensors, acceleration sensors, vibration sensors, and optical sensors, etc. Data collection starts immediately after the maintenance is completed and lasts for at least a 24-hour operation cycle to ensure coverage of various working states of the strapping machine. The collected raw data undergoes the same filtering, segmentation processing, and feature extraction processes as in the initial monitoring to form a standardized performance index data set. The performance indexes include key parameters such as motor temperature stability, tension control accuracy, robotic arm movement smoothness, cutting vibration spectrum characteristics, and strapping deformation consistency. These index data are sorted in a unified format to form a structured evaluation report on the operation effect after maintenance, which contains the numerical values of various performance indexes, the time series change trends, and statistical characteristics. The evaluation report on the operation effect after maintenance is compared and analyzed with the state before maintenance to calculate the improvement measurement value of the maintenance effect. The comparative analysis uses the parameter difference method, comparing each performance index before and after maintenance one by one to calculate the relative change amount and the absolute change amount. The relative change amount represents the percentage of index improvement, and the absolute change amount represents the actual changed value of the index. For each performance index, a standardized improvement index is calculated to unify various indexes with different dimensions into the range of 0 - 1. Then, the improvement indexes of each index are weighted and averaged according to their importance to form a comprehensive improvement measurement value of the maintenance effect. This measurement value intuitively reflects the degree of improvement in the overall performance of the strapping machine by the maintenance activity, with a numerical range of 0 - 1. The closer the value is to 1, the more significant the improvement effect.

[0107] Compare the fault cause location results with the actual fault sources found during maintenance, and record the evaluation data on the accuracy of fault diagnosis. During the implementation of the maintenance plan, maintenance personnel will discover the actual fault sources, which are detailedly recorded in the maintenance record form, including the faulty components, fault types, fault severity, and possible fault causes. By comparing the fault cause location results given by the automatic diagnosis system with the actual fault sources found by the maintenance personnel, the accuracy of the diagnosis is evaluated. The evaluation indicators include: the matching degree of the target faulty component (the consistency between the faulty component diagnosed automatically and the actual faulty component), the matching degree of the fault type (the consistency between the fault type diagnosed automatically and the actual fault type), and the deviation of the estimated fault severity (the difference between the predicted severity and the actual severity). These matching degrees and deviation values constitute the evaluation data on the accuracy of fault diagnosis. According to the deviation situation in the evaluation data on the accuracy of fault diagnosis, adjust the feature thresholds of the corresponding fault modes in the fault feature library of the strapping machine. The fault feature library is a set of feature templates for fault identification, and each fault mode has a corresponding feature threshold setting. The adjustment process adopts an incremental learning strategy, and the feature thresholds are fine-tuned according to the deviation situation of the current diagnosis. For misdiagnosis cases, analyze the key feature parameters that lead to misdiagnosis, and appropriately adjust their weights and threshold ranges; for missed diagnosis cases, enhance the sensitivity of relevant feature parameters and expand the coverage range of the feature thresholds; for cases where the diagnosis is correct but there is a deviation in the estimated severity, correct the coefficients in the severity evaluation model. The adjusted feature thresholds will be immediately updated to the fault feature library for the next fault diagnosis.

[0108] Update the timing parameters in the prediction algorithm according to the differences between the predicted values of the fault development trend and the actual fault development trajectory. The fault development trend prediction algorithm includes multiple timing parameters, such as the development rate coefficient, the environmental impact factor, the load adjustment coefficient, etc. Calculate the prediction deviation by comparing the predicted fault development trend with the actual fault development observed during the maintenance process. The deviation analysis includes: the development speed deviation (the difference between the predicted speed and the actual speed), the inflection point time deviation (the time difference between the predicted inflection point and the actual inflection point), and the severity deviation (the difference between the predicted severity and the actual severity). According to these deviations, use the gradient descent method to iteratively optimize the timing parameters in the prediction algorithm to reduce the prediction error. The optimization process will consider the historical correction records to avoid over-adjustment caused by single samples, and ensure the stability and accuracy of the prediction model are gradually improved. Integrate the associated data between the maintenance effect improvement metric and the maintenance plan and write it into the maintenance knowledge parameter library to form an updated fault feature library and maintenance knowledge parameters for the belt machine. The maintenance knowledge parameter library is a structured database that stores maintenance experience and best practices, and is used to guide future maintenance decisions. The integration process associates the key information of the current maintenance case (fault type, maintenance method, resources used, maintenance duration, etc.) with the maintenance effect improvement metric to form a complete maintenance record; then mark the effectiveness level of the maintenance plan according to the quality of the maintenance effect; then write the integrated data into the maintenance knowledge parameter library in a predetermined format; update the statistical indicators in the maintenance knowledge parameter library, such as the average maintenance time and average maintenance effect of various faults. At the same time, update the adjusted fault feature threshold and the optimized prediction algorithm parameters to the fault feature library and the prediction model respectively to form a complete knowledge update cycle.

[0109] For example, after a certain banding machine completed the replacement and repair of the cutting tool, the sensor network immediately started collecting performance data. The motor temperature sensor showed that the temperature dropped from the previous continuous high temperature (85°C) to the normal range (70°C); the abnormal peak at 150 Hz in the vibration spectrum recorded by the vibration sensor during the cutting process disappeared, and the vibration amplitude decreased; the noise sensor detected that the cutting noise dropped from 78 dB to the standard level of 65 dB; the quality parameters of the banding incision measured by the optical sensor returned to normal. After these data were standardized, an evaluation report on the operation effect after the repair was formed. Comparative analysis with the state before the repair showed that the motor temperature decreased by 15°C (improvement degree 0.88), the abnormal vibration spectrum disappeared (improvement degree 0.95), the noise level decreased by 13 dB (improvement degree 0.90), and the comprehensive calculated improvement measure value of the repair effect was 0.92, indicating that the repair was very successful. The actual findings of the maintenance personnel confirmed that the serious wear of the cutting tool was the root cause of the failure, which was completely consistent with the result of "cutting tool wear" given by the automatic diagnosis system. The matching degree of the target fault component was 1.0, and the matching degree of the fault type was 1.0. However, in terms of the severity of the fault, the severity estimated by the automatic system was 0.75, while the actually found severity was 0.85, with an underestimation deviation of 0.1. Accordingly, the feature threshold of the cutting tool wear mode in the fault feature library was adjusted, especially the mapping relationship between the energy at 150 Hz in the vibration spectrum and the severity of the fault was enhanced, and the severity evaluation model was corrected. At the same time, by comparing the predicted fault development trend with the actual observation results, it was found that the predicted development speed of the system was 15% slower than the actual one. Accordingly, the development rate coefficient in the prediction algorithm was updated from the original 0.085 to 0.098 to make the prediction closer to the actual situation. The complete information of this repair (including fault diagnosis, repair method, resources used, repair duration of 2 hours, and repair effect improvement measure value of 0.92, etc.) was written into the maintenance knowledge parameter library, and at the same time, statistical indicators such as the average repair effect and average repair duration of cutting tool replacement were updated.

[0110] The above described the method for monitoring the faults of the banding machine based on the Internet of Things in the embodiments of the present application. Next, the system for monitoring the faults of the banding machine based on the Internet of Things in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the system for monitoring the faults of the banding machine based on the Internet of Things in the embodiments of the present application includes:

[0111] An extraction module, configured to collect the operation data of each key part of the banding machine through the sensor network, and obtain multi-source operation data including motor temperature data, tension control parameters, displacement data of the bundling robotic arm, operation status of the cutting device, and banding deformation rate;

[0112] A segmentation module, configured to perform filtering and segmentation processing on the data according to the multi-source operation data, and obtain the state feature data of the banding machine with time stamps;

[0113] A comparison module, configured to analyze and compare the operating status of the bundling machine based on the state characteristic data of the bundling machine, and obtain an abnormal status identifier of the bundling machine and a determination result of the fault type;

[0114] A positioning module, configured to establish an associated graph of bundling machine components according to the abnormal status identifier of the bundling machine and the determination result of the fault type, and obtain a fault cause positioning result and a predicted value of the fault development trend;

[0115] A calculation module, configured to calculate the maintenance priority in combination with the bundling machine maintenance database according to the fault cause positioning result and the predicted value of the fault development trend, and obtain a maintenance plan and an operation time suggestion;

[0116] A prediction module, configured to record the actual maintenance result and the prediction deviation according to the operating effect of the bundling machine after the implementation of the maintenance plan, and obtain an updated bundling machine fault feature library and maintenance knowledge parameters.

[0117] Through the collaborative cooperation of the above-mentioned various components,

[0118] Referring to Figure 3 , in an embodiment of the present invention, a computer device is further provided. The computer device may be a server, and its internal structure may be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network. When the computer program is executed by the processor, the above method is implemented.

[0119] Those skilled in the art can understand that Figure 3 the structure shown in

[0120] is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0121] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0122] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0123] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs and other various media that can store program codes.

[0124] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring faults of a strapping machine based on the Internet of Things, characterized in that: include: The sensor network collects the operating data of key parts of the strapping machine, and obtains multi-source operating data including motor temperature data, tension control parameters, strapping robot arm displacement data, cutting device operating status and strap deformation rate; According to the multi-source operation data, the data is filtered and segmented to obtain the state characteristic data of the strapping machine with time stamp; Based on the state characteristic data of the strapping machine, the running state of the strapping machine is analyzed and compared to obtain the abnormal state identification and fault type determination results of the strapping machine; According to the abnormal state identification and fault type determination results of the strapping machine, a component association map of the strapping machine is established to obtain the fault cause location results and the fault development trend prediction value; According to the fault cause location results and the predicted value of fault development trend, the maintenance priority is calculated in combination with the belt machine maintenance database to obtain the maintenance plan and operation time recommendations; According to the running effect of the strapping machine after the maintenance plan is implemented, the actual maintenance results and the predicted deviation are recorded to obtain the updated strapping machine fault feature library and maintenance knowledge parameters; The method includes collecting the performance indicator data of the belting machine after the implementation of the maintenance plan through the sensor network, and constructing a post-maintenance operation effect evaluation report; comparing and analyzing the post-maintenance operation effect evaluation report with the state before maintenance, and calculating the maintenance effect improvement metric; comparing the fault cause location result with the fault source found in the actual maintenance, and recording the fault diagnosis accuracy evaluation data; adjusting the feature threshold of the corresponding fault mode in the fault feature library of the belting machine according to the deviation in the fault diagnosis accuracy evaluation data, adjusting the incremental learning strategy, and fine-tuning the feature threshold according to the deviation of the current diagnosis. For misdiagnosis, analyze the key feature parameters that cause misdiagnosis, and appropriately adjust their weights and threshold ranges; for missed diagnosis, enhance the sensitivity of related feature parameters and expand the coverage of feature thresholds; for correct diagnosis but deviation in severity estimation, correct the coefficients in the severity evaluation model; according to the difference between the predicted value of the fault development trend and the actual fault development trajectory, update the timing parameters in the prediction algorithm; integrate the maintenance effect improvement metric and the associated data of the maintenance plan into the maintenance knowledge parameter library to form an updated belting machine fault feature library and maintenance knowledge parameters.

2. The method for monitoring faults of a strapping machine based on the Internet of Things according to claim 1 is characterized in that: The sensor network collects the operating data of key parts of the strapping machine, and obtains multi-source operating data including motor temperature data, tension control parameters, strapping robot arm displacement data, cutting device operating status and strap deformation rate, including: Install temperature sensors and current sensors on the motor drive device of the strapping machine to collect the motor surface temperature and current waveform to obtain the motor temperature data; Arrange tension sensors and displacement sensors in the tension control mechanism of the strapping machine to record the tension change of the strapping machine and the displacement of the control rod to form tension control parameters; The binding robot arm of the strapping machine is equipped with an acceleration sensor and an angle sensor to measure the movement state and working angle of the robot arm and generate the displacement data of the binding robot arm; Install vibration sensors and sound sensors on the cutting device of the strapping machine to monitor the vibration frequency and noise characteristics during the cutting process and generate the operating status of the cutting device; The deformation of the strap during the bundling process is measured by an optical sensor, the degree of deformation of the strap due to stress is calculated, and the deformation rate of the strap is obtained; The motor temperature data, tension control parameters, tying robot arm displacement data, cutting device operating status and belt deformation rate are transmitted to the data processing unit through an industrial-grade wireless communication network to form multi-source operating data with corresponding time stamps.

3. The method for monitoring faults of a strapping machine based on the Internet of Things according to claim 2 is characterized in that: According to the multi-source operation data, the data is filtered and segmented to obtain the state characteristic data of the belting machine with time stamp, including: The noise and interference signals in the multi-source operation data are removed through adaptive filtering algorithms to obtain a pure data stream; The pure data stream is divided into data segments according to the sliding window method with a width of 5 seconds and a 50% overlap rate to form segmented processing data; The time synchronization mechanism adds precise time tags to the segmented processed data to construct time-aligned data frames; Extracting time domain features including mean, peak, root mean square value, kurtosis and skewness from the time aligned data frame by time domain analysis method; Perform frequency domain conversion on the time-aligned data frame to obtain spectrum energy distribution and main frequency component information; The time domain feature quantity is fused with the spectrum energy distribution and the main frequency component information to generate the state feature data of the belting machine with timestamp.

4. The method for monitoring faults of a strapping machine based on the Internet of Things according to claim 3 is characterized in that: Based on the state characteristic data of the strapping machine, the running state of the strapping machine is analyzed and compared to obtain the abnormal state identification and fault type determination results of the strapping machine, including: Construct a normal operating benchmark parameter library of the belt machine from historical data, including the standard characteristic distribution boundary values ​​under various working conditions; The difference between the state characteristic data of the strapping machine and the corresponding data in the normal operation benchmark parameter library of the strapping machine is calculated to obtain the characteristic deviation matrix; The anomaly weight coefficient is set according to the degree of deviation of each parameter in the characteristic deviation matrix to generate a comprehensive anomaly index; The threshold comparison method is used to determine whether the comprehensive abnormal index exceeds the preset warning threshold, forming an abnormal state mark of the strapping machine; According to the typical fault feature template of the strapping machine, similarity matching is performed on the feature deviation matrix to identify the target fault category; The abnormal state identification of the strapping machine and the target fault category are combined to form the abnormal state identification of the strapping machine and the fault type determination result.

5. The method for monitoring faults of a strapping machine based on the Internet of Things according to claim 4 is characterized in that: According to the abnormal state identification and fault type determination results of the strapping machine, a component association map of the strapping machine is established to obtain the fault cause location results and the fault development trend prediction value, including: Using the structural knowledge of the strapping machine, a component association graph of the strapping machine is constructed, which includes the physical connections and functional dependencies between the components. Mark abnormal component nodes in the belt machine component association graph according to the fault type determination result; The abnormal signal transmission path is traced back in the belting machine component association map through the causal propagation algorithm, and the probability value of each related component as the root cause of the fault is calculated; Determine the target faulty components and their impact range by sorting the probability values, and generate the fault cause location results; Based on the development rules of similar historical fault cases, the fault cause location results are extrapolated in time series to construct a fault severity change curve; By combining the fault severity change curve with the equipment safety threshold, the fault development time window and potential risk level are predicted to form a fault development trend prediction value.

6. The method for monitoring faults of a strapping machine based on the Internet of Things according to claim 5 is characterized in that: Based on the fault cause location results and the predicted value of the fault development trend, the maintenance priority is calculated in combination with the belt machine maintenance database to obtain maintenance plans and operation time recommendations, including: Extract historical maintenance cases that match the fault cause location results from the belt machine maintenance database to build a basic set of maintenance knowledge; Predicting the fault development trend according to the maintenance knowledge base set, obtaining the fault development trend prediction value, and calculating the fault risk index according to the fault development trend prediction value; The failure risk index is weighted and combined with the production task priority, spare parts inventory status and technician availability to generate a maintenance urgency index; Classify maintenance tasks through maintenance urgency index and determine maintenance operation level; Extract the corresponding standard maintenance process based on the maintenance knowledge base set and maintenance operation level to form a maintenance plan; Combine the predicted value of fault development trend with the production plan to determine the optimal maintenance time window and generate operation time recommendations.

7. A strapping machine fault monitoring system based on the Internet of Things, used to implement the strapping machine fault monitoring method based on the Internet of Things as described in any one of claims 1 to 6, characterized in that: The fault monitoring system of the strapping machine based on the Internet of Things includes: The extraction module is used to collect the operating data of each key part of the strapping machine through the sensor network, and obtain multi-source operating data including motor temperature data, tension control parameters, strapping robot arm displacement data, cutting device operating status and strap deformation rate; The segmentation module is used to filter and segment the data according to the multi-source operation data to obtain the state characteristic data of the strapping machine with a timestamp; A comparison module is used to analyze and compare the running state of the strapping machine based on the state characteristic data of the strapping machine, and obtain the abnormal state identification and fault type determination result of the strapping machine; The positioning module is used to establish a correlation map of the parts of the strapping machine according to the abnormal state identification and fault type determination results of the strapping machine, and obtain the fault cause positioning results and the fault development trend prediction value; The calculation module is used to calculate the maintenance priority based on the fault cause location results and the fault development trend prediction value, combined with the belt machine maintenance database, to obtain the maintenance plan and operation time recommendation; The prediction module is used to record the actual maintenance results and the predicted deviation according to the operation effect of the belting machine after the maintenance plan is implemented, and obtain the updated fault feature library and maintenance knowledge parameters of the belting machine.

8. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the method for monitoring the fault of a strapping machine based on the Internet of Things described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor executes the method for monitoring faults of a strapping machine based on the Internet of Things as claimed in any one of claims 1 to 6.

Images