AI-based Internet of Things equipment terminal health state pre-sensing analysis method

Abstract

The invention discloses an Internet of Things equipment terminal health state pre-sensing analysis method based on AI, and relates to the technical field of equipment health management.The method comprises the steps that multiple types of sensors are deployed on key components to collect operation data, response data sets are formed through preprocessing, and a response data set set Ai is formed through set processing; calculating a health state scoring index hss to perform operation health assessment, if the operation state is normal, constructing a wear prediction model through a convolutional neural network model to obtain an aging wear prediction index wap, generating an aging influence propagation coefficient wip, and calculating an aging response propagation index fpx in combination with the health state scoring index hss and the aging wear prediction index wap; and finally, the fault propagation probabilities of all the components are accumulated, a comprehensive fault risk prediction index SRR is obtained for comprehensive fault response evaluation, real-time evaluation and early warning of the operation state of the industrial robot are achieved, stable operation of equipment is ensured, and the service life of the equipment is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment health management, and particularly to a method for pre-perceiving and analyzing the health status of IoT device terminals based on AI. Background Art

[0002] With the deep promotion of intelligent manufacturing, industrial intelligence and digital transformation, society's dependence on automated equipment with high efficiency, stability and safe operation is increasing day by day. Especially in key fields such as automobile manufacturing, semiconductor assembly, and high-end equipment, the requirements for production continuity and equipment reliability are constantly improving. At the same time, the rising labor cost and the shortage of high-skilled talents also prompt enterprises to increasingly rely on intelligent systems to replace the traditional manual operation and maintenance mode. Facing the strong demand for intelligent perception and prediction capabilities of equipment operation in industrial upgrading, promoting the transformation of industrial equipment from "post-failure repair" to "early warning" has become an inevitable trend in the digital development of the manufacturing industry.

[0003] At present, the health status monitoring methods of industrial robots generally rely on regular maintenance or a single-parameter threshold-triggered warning mechanism. This method has significant lag and locality problems, and it is difficult to capture the potential wear behavior and the gradual deterioration process at the equipment system level. At the same time, most of the existing methods are based on static parameters such as conventional voltage, current or temperature for judgment, lacking a dynamic understanding of the equipment state evolution process, and it is not easy to make a systematic evaluation of the associated effects of aging propagation between components. In addition, the current technology also faces challenges at the data processing level. For example, it is not easy to effectively extract the implicit features strongly related to the health status from high-dimensional collected data, and there is a lack of a systematic filtering mechanism for abnormal data and noise data, resulting in model training deviation and unstable prediction accuracy. Therefore, there is an urgent need for a systematic method that integrates multi-dimensional time-series signal processing, aging propagation modeling and component-level functional coupling analysis to comprehensively improve the scientificity and reliability of industrial robot health status monitoring and prediction. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method for pre-perceiving and analyzing the health status of IoT device terminals based on AI, which solves the problems in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for pre-perceiving and analyzing the health status of IoT device terminals based on AI, including the following steps: S1. According to the sensor group installed on the industrial robot, the operation data of the industrial robot is collected in real time and transmitted to the local server for preprocessing to obtain a response data group; S2. The local server performs set processing on the response data group to form a response data group set A i and, according to the response data group set A iCalculate the health status scoring index hss for running health assessment; S3. When the running health assessment indicates that the device is operating normally, construct a wear prediction model based on the convolutional neural network CNN, and then input the obtained set of response data groups A i into the wear prediction model to obtain the aging wear prediction index wap; S4. The local server defines the functional relationships of components based on the electronic structure drawings and component working link diagrams of the industrial robot, and uses the historical collaborative fault data learning method to generate the aging influence propagation coefficient w ip , and then combines the health status scoring index hss and the aging wear prediction index wap to calculate the aging response propagation index fpx; S5. The local server non-weightedly accumulates the fault propagation probability values of all components of the industrial robot based on the obtained aging response propagation index fpx, and calculates the comprehensive fault risk prediction index SRR for comprehensive fault response assessment.

[0006] Preferably, S1 includes S11 and S12; S11. Install a sensor group on each key component of the industrial robot, and collect the running data of the industrial robot in real time through the installed sensor group; The sensor group includes a temperature sensor, a piezoelectric sensor, a vibration sensor, a current sensor, a voltage sensor, and a humidity sensor.

[0007] Preferably, S12. Transmit the running data collected by the sensor group to the local server through a wireless network. The local server extracts the feature data related to the health status of the industrial robot from the running data through principal component analysis technology; The principal component analysis technology extracts the feature data related to the health status of the industrial robot from multiple original features of the running data through dimensionality reduction; After the local server obtains the feature data, it performs data cleaning, denoising, outlier detection, and dimensionless processing on the feature data to obtain a set of response data; Data cleaning automatically eliminates outlier data through the set data quality threshold. Denoising weights and fuses historical data with current feature data through the Kalman filter algorithm to filter the noise influence in the feature data. Outlier detection identifies abnormal patterns in the running data through a machine learning model combined with historical data. Dimensionless processing eliminates the dimensional influence of the running data through the Max-Min maximum-minimum method; The set of response data includes the running temperature wd, running load fz, vibration amplitude zf, current intensity dl, voltage intensity yq, and humidity sd.

[0008] Preferably, S2 includes S21; S21. The local server aggregates the response data groups of the same i-component of the industrial robot collected at time t to form a set of response data groups A i , denoted as: A i ⊇ {wd, fz, zf, dl, yq, sd}, and after marking the time stamp, it is stored in the data repository.

[0009] Preferably, S2 further includes S22; S22. Analyze the health status of each component of the industrial robot based on the obtained set of response data groups A i , and then analyze the health status of the industrial robot and conduct a health assessment on the industrial robot, specifically including S221 and S222; S221. The local server calculates and obtains the health status score index hss based on set A i , and the specific calculation formula is as follows; ; In the formula, hss i (t) represents the health status score index of the i-th component at time t, m represents the total number of collected parameters, k i，j represents the adjustment factor of the j-th parameter to the health status score index of the i-th component, A i，j (t) represents the value of the j-th parameter collected in real time by the i-th component at time t, B i，j represents the standard value of the j-th parameter of the i-th component under standard working conditions, λ j represents the time decay factor of the j-th parameter, exp represents the exponential decay function, t represents the current time t, and t0 represents the initial time when the i-th component starts to work.

[0010] Preferably, S222. Based on the health operation industry standard in the field of industrial robot equipment health management, a preset mechanical operation health threshold X is set, and then compared with the obtained health status score index hss to conduct an operation health assessment. The specific assessment plan is as follows; When the health status score index hss ≤ the mechanical operation health threshold X, it indicates that the equipment has a fault. At this time, according to the robot control system, the machine is immediately stopped, and a fault message is generated and transmitted to the relevant personnel's user terminal through wireless communication to remind that the machine fault requires maintenance; When the health status score index hss > the mechanical operation health threshold X, it indicates that the equipment is operating normally. At this time, the equipment prediction instruction is executed.

[0011] Preferably, S3 includes S31 and S32; S31. When the running health assessment indicates that the device is operating normally, the local server constructs a wear prediction model based on the convolutional neural network CNN and queries the set A of time series response data groups of historical industrial robot devices in the data repository. i Analyze the timing characteristics of the historical response data groups, then extract the frequency components in the device vibration signal through the fast Fourier transform, analyze the changes of the historical response data groups in different frequency ranges, obtain the historical timing characteristic response data set, and input the historical timing characteristic response data set into the wear prediction model for model training to optimize the wear prediction model. S32. Input the set A of response data groups obtained in real time i into the trained wear prediction model for device aging prediction. Output the aging wear prediction index wap through the wear prediction model. The specific formula for analyzing the aging and wear of the industrial robot device is as follows; ; In the formula, wap i (t) represents the aging wear prediction index of the i-th component at time t, m represents the total number of collected parameters, and α i represents the adjustment coefficient of the wear rate of the i-th component, and α i，0 represents the initial wear coefficient of the i-th component, e represents the exponential function, and γ i represents the time decay coefficient of the i-th component.

[0012] Preferably, S4 includes S41 and S42; S41. After performing device aging prediction, the local server defines the functional dependencies and physical conduction relationships between the various components of the industrial robot based on the electronic structure drawings and component working link diagrams of the industrial robot, uses the system modeling language and the BOM structure list to identify the hierarchical and coupling relationships between the device modules, and generates the aging influence propagation coefficient w using the historical co-failure data learning method ip , specifically as follows; ; w ip represents the causal intensity of the failure of component i on component p; S42. The local server performs comprehensive calculations based on the obtained health status score index hss, aging wear prediction index wap, and aging influence propagation coefficient w ip to obtain the aging response propagation index fpx and analyze the impact of the aging of a single component on the operation of the entire robot; ; In the formula, fpx p (t) represents the failure response propagation index of the aging component on the p-th component at time t, N represents the total number of robot components, and e represents the exponential function.

[0013] Preferably, S5 includes S51; S51. The local server non - weighted accumulates the failure propagation probability values of all components of the industrial robot according to the obtained aging response propagation index fpx, calculates and obtains the comprehensive failure risk prediction index SRR, and analyzes the overall failure risk of the industrial robot. The specific formula is as follows; ; In the formula, (t) represents the mean value of the aging response propagation indices of all components at time t.

[0014] Preferably, S5 also includes S52; S52. Collect the comprehensive failure risk prediction indices SRR of the industrial robot in normal and faulty states in each historical period, calculate the mean value of the historical comprehensive failure risk prediction index SRR according to the statistical method, and set a preset operation failure response threshold V based on the mean value. Then compare it with the obtained comprehensive failure risk prediction index SRR to conduct a comprehensive failure response assessment of the industrial robot. The specific assessment scheme is as follows; When the comprehensive failure risk prediction index SRR ≥ operation failure response threshold V, it indicates that there are aging components in the industrial robot, and the aging components affect the normal operation of other components. At this time, control the robot to stop immediately according to the robot control system, and generate aging component information and transmit it to the relevant personnel's user terminal through wireless communication to remind to replace the aging equipment components; When the comprehensive failure risk prediction index SRR < operation failure response threshold V, it indicates that the robot components are normal, and normal monitoring is maintained.

[0015] The present invention provides a method for pre - sensing and analyzing the health status of the IoT device terminal based on AI. It has the following beneficial effects: (1) By installing a variety of sensors, this method realizes the comprehensive collection of the operation data of the key components of the industrial robot, and through the principal component analysis method, extracts and pre - processes the data features, obtains a response data group highly correlated with the equipment health status, provides a data basis for subsequent analysis, and constructs a response data group set A according to the response data group i , and by calculating the health status scoring index hss, quantifies the current health level of the equipment, and based on the set mechanical operation health threshold X, realizes the preliminary assessment and fault determination of the equipment operation status.

[0016] (2) This method constructs a wear prediction model based on the convolutional neural network CNN, introduces Fourier transform to extract vibration frequency features, combines historical response data sets to train and optimize the model, so that it outputs the aging wear prediction index wap, comprehensively evaluating the wear trend of the device under the influence of different parameters. Immediately afterwards, the electronic structure diagram of the device and the component link structure are introduced to construct the aging propagation path between components, and the aging influence propagation coefficient w is established through the historical collaborative fault data learning method ip , thereby further calculating the aging response propagation index fpx, effectively revealing the potential impact of the aging of a certain component on other modules, and providing a theoretical basis for accurate prediction.

[0017] (3) This method non-weightedly accumulates the aging response propagation indexes fpx of all components to form a comprehensive fault risk prediction index SRR, and sets an operating fault response threshold V in combination with historical operation samples to complete the quantitative judgment of device-level fault response. This method can identify the aging trend in advance and trigger an alarm before the device actually fails, guide the adjustment and implementation of the device maintenance strategy, avoid fault diffusion and system shutdown, effectively extend the service life of the device, and improve the stability and operation and maintenance efficiency of industrial robots. The overall solution realizes the full-closed-loop health management process from perception, evaluation, prediction to response, and is an important support for the integrated development of intelligent manufacturing and device status perception. Description of the Drawings

[0018] Figure 1 is a schematic diagram of the steps of the method for pre-perceiving and analyzing the health status of the IoT device terminal based on AI of the present invention; Figure 2 is a schematic diagram of the line graph representing the comprehensive fault response evaluation of the present invention; Figure 3 is a schematic diagram of the process framework of the method for pre-perceiving and analyzing the health status of the IoT device terminal based on AI of the present invention. Specific Embodiments

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment 1

[0021] Please refer to Figure 1 , the present invention provides a method for pre-perceiving and analyzing the health status of the IoT device terminal based on AI. To achieve the above objectives, the present invention is realized through the following technical solutions: including the following steps: S1. Based on the sensor group installed on the industrial robot, the operation data of the industrial robot is collected in real time and transmitted to the local server for preprocessing to obtain a response data group; S2. The local server performs set processing on the response data group to form a response data group set A i , and calculates the health status scoring index hss based on the response data group set A i for operation health assessment; S3. When the operation health assessment indicates that the equipment is operating normally, a wear prediction model is constructed based on the convolutional neural network CNN, and then the obtained response data group set A i is input into the wear prediction model to obtain the aging wear prediction index wap; S4. The local server defines the functional relationship of components based on the electronic structure drawings and component working link diagrams of the industrial robot, and uses the historical collaborative fault data learning method to generate the aging influence propagation coefficient w ip , and then combines the health status scoring index hss and the aging wear prediction index wap to calculate the aging response propagation index fpx; S5. The local server non-weightedly accumulates the fault propagation probability values of all components of the industrial robot based on the obtained aging response propagation index fpx, and calculates the comprehensive fault risk prediction index SRR for comprehensive fault response assessment.

[0022] In this embodiment, S1 realizes high-frequency and multi-dimensional real-time data collection of the operation status of each key component of the industrial robot through various types of sensor groups, and preprocesses it through the local server to form a highly purified response data group, laying a solid foundation for subsequent intelligent processing. This stage significantly improves the timeliness and effectiveness of operation data, and solves the problems of messy sensor data and difficult real-time processing in traditional solutions. In the S2 and S3 stages, the local server calculates the health status scoring index hss based on the response data group set A i , and further judges whether the equipment is in a healthy state. Once it is judged that the equipment is operating normally, it automatically enters the wear prediction process based on the convolutional neural network CNN. Through the time series modeling and frequency analysis of historical data, the aging wear prediction index wap is generated. Compared with the traditional passive strategy that relies on periodic manual maintenance, this method can actively predict the wear trend of components, greatly improve the forward-looking and pertinence of equipment maintenance, reduce the sudden failure rate of equipment, and realize the transformation from "fault response" to "wear prediction". In the S4 and S5 stages, further define the functional dependence and physical conduction relationship between each component of the industrial robot based on the electronic structure drawings and component working link diagrams of the industrial robot, use the system modeling language and BOM structure list to identify the hierarchy and coupling relationship between equipment modules, and use the historical collaborative fault data learning method to construct the aging propagation influence coefficient w between components ip, and generate the aging response propagation index fpx accordingly. Finally, the propagation indices of all components are cumulatively added without weighting to obtain the comprehensive fault risk prediction index SRR, realizing the macroscopic risk assessment of the overall operation status of the machine. Compared with the current industry solutions mainly based on single-parameter indicators or simple alarm mechanisms, this method introduces a response propagation analysis mechanism that integrates multiple parameters and multiple models, effectively improving the accuracy of risk assessment and the system's intelligent decision-making ability, and further promoting the industrial robot to move towards a highly reliable, highly autonomous, and high-efficiency operation and management mode.

[0023] Embodiment 2

[0024] This embodiment is an explanatory description carried out in Embodiment 1, please refer to Figure 1 , specifically: S1 includes S11 and S12; S11. Install sensor groups on each key component of the industrial robot, and collect the operation data of the industrial robot in real time through the installed sensor groups; The sensor group includes a temperature sensor, a piezoelectric sensor, a vibration sensor, a current sensor, a voltage sensor, and a humidity sensor; The temperature sensor is used to collect the operating temperature of the industrial robot in real time; The piezoelectric sensor is used to collect the operating load of the industrial robot in real time; The vibration sensor is used to collect the vibration amplitude of the equipment during the operation of the industrial robot in real time; The current sensor is used to detect the current intensity of the industrial robot in real time; The voltage sensor is used to detect the voltage intensity of the industrial robot in real time; The humidity sensor is used to collect the ambient humidity in real time.

[0025] S12. Transmit the operation data collected by the sensor group to the local server through a wireless network. The local server extracts the feature data related to the health status of the industrial robot from the operation data through principal component analysis technology; The principal component analysis technology extracts the feature data related to the health status of the industrial robot from multiple original features of the operation data by means of dimensionality reduction; After obtaining the feature data, the local server performs data cleaning, denoising, outlier detection, and dimensionless processing on the feature data to obtain a response data group; Data cleaning automatically eliminates outlier data through the set data quality threshold. Denoising weights and fuses historical data with current feature data through the Kalman filter algorithm to filter the noise influence in the feature data. Outlier detection identifies abnormal patterns in the operation data through a machine learning model combined with historical data. Dimensionless processing eliminates the dimensional influence of the operation data through the Max-Min maximum-minimum method; The response data set includes the operating temperature wd, the operating load fz, the vibration amplitude zf, the current intensity dl, the voltage intensity yq, and the humidity sd.

[0026] In this embodiment, by deploying a sensor group on the key components of the industrial robot, the real-time acquisition of multi-dimensional parameters of the device operation data is realized, and the data is transmitted to the local server through the wireless network; the local server extracts the feature factors highly related to the health state from the high-dimensional operation data based on the principal component analysis technology, and combines multiple data optimization means such as data cleaning, Kalman filter denoising, outlier identification, and dimensionless processing to form a response data set with clear structure, low noise, and discriminant value. This method not only improves the accuracy and timeliness of the operation data processing, but also effectively avoids the problems of data redundancy, weak feature correlation, and evaluation lag in the traditional method, significantly improving the reliability and response speed of the subsequent health state evaluation and aging prediction model, and providing a solid foundation for building an intelligent and predictable robot operation and maintenance system.

[0027] Embodiment 3

[0028] This embodiment is an explanatory description based on Embodiment 2. Please refer to Figure 1 , specifically: S2 includes S21 and S22; S21. The local server aggregates the response data sets of the same i-th component of the industrial robot collected at time t to form a response data set collection A i , denoted as: A i ⊇{wd, fz, zf, dl, yq, sd}, and after marking the time stamp, it is stored in the data repository.

[0029] S22. Based on the obtained response data set collection A i Analyze the health state of each component of the industrial robot, and then analyze the health state of the industrial robot and conduct a health assessment on the industrial robot, specifically including S221 and S222; S221. The local server calculates and obtains the health state scoring index hss based on the set A i , and the specific calculation formula is as follows; ; In the formula, hss i (t) represents the health state scoring index of the i-th component at time t, m represents the total number of collected parameters, and k i，j represents the adjustment factor of the j-th parameter to the health state scoring index of the i-th component, which is used to adjust the sensitivity and importance of the parameter to the health state scoring. A i，j (t) represents the value of the j-th parameter collected in real time by the i-th component at time t, and B i，jrepresents the standard value of the j-th parameter of the i-th component under standard working conditions, λ j represents the time decay factor of the j-th parameter, reflecting the decay of the influence of the j-th parameter on the health state over time, which is extracted through the experimental data management system of the material supplier. exp represents the exponential decay function, t represents the current time t, and t0 represents the initial time when the i-th component starts working. represents the term of the j-th parameter's effect on the health state at the current time t, that is, the degree of deviation of this parameter. represents the time decay function of the j-th parameter, describing the natural decay and boundary decreasing effect of the j-th parameter on the health state scoring index over time.

[0030] S222. Based on the health operation industry standard in the field of industrial robot equipment health management, a preset mechanical operation health threshold X is set, and then compared with the obtained health state scoring index hss to conduct an operation health assessment. The specific assessment scheme is as follows; When the health state scoring index hss ≤ the mechanical operation health threshold X, it indicates that the equipment has a fault. At this time, the robot control system controls the machine to stop immediately, and generates a fault message and transmits the fault message to the relevant personnel's user terminal through wireless communication to remind that the machine fault needs to be repaired; When the health state scoring index hss > the mechanical operation health threshold X, it indicates that the equipment is operating normally. At this time, the equipment prediction instruction is executed.

[0031] In this embodiment, S21 forms a standardized response data group set A by aggregating six types of key response parameters of the same component at the same time i , and stores it with a time stamp to ensure the temporal integrity of the data; in S22, further by introducing an adjustment factor k i，j , the time decay factor λ j and the standard working condition parameter B i，j , calculate the health state scoring index hss, accurately quantify the health state of each component, and make a judgment in combination with the preset mechanical operation health threshold X in the industry, realizing real-time, dynamic and quantitative assessment of the operation state of industrial robots. This method not only breaks through the traditional health assessment method based on single parameter or manual experience, realizes the early perception and rapid response to equipment faults, but also improves the reliability, safety and operation and maintenance efficiency of robot operation, laying a core technical foundation for building an intelligent predictive maintenance system.

[0032] Embodiment 4

[0033] This embodiment is an explanatory description carried out in Embodiment 3. Please refer to Figure 1 , specifically: S3 includes S31 and S32; S31. When the running health assessment indicates that the device is operating normally, the local server constructs a wear prediction model based on the convolutional neural network (CNN), and queries the set A of time series response data groups of historical industrial robot devices in the data repository. i Analyze the timing characteristics of the historical response data groups, then extract the frequency components in the device vibration signal through fast Fourier transform, analyze the changes of the historical response data groups in different frequency ranges, obtain the historical timing feature response data set, and input the historical timing feature response data set into the wear prediction model for model training to optimize the wear prediction model. S32. Input the set A of response data groups obtained in real time i into the trained wear prediction model for equipment aging prediction. Output the aging wear prediction index wap through the wear prediction model. The specific formula for analyzing the aging wear of the industrial robot device is as follows; ; In the formula, wap i (t) represents the aging wear prediction index of the i-th component at time t, which is used to describe the wear and aging conditions of the i-th component at time t. m represents the total number of collected parameters, and α i represents the adjustment coefficient of the wear rate of the i-th component, which describes the wear degree of the component under different factors and is extracted through the experimental data management system of the material supplier. α i，0 represents the initial wear coefficient of the i-th component, which is extracted through the material management system of the material supplier. e represents the exponential function, and γ i represents the time decay coefficient of the i-th component, which is used to describe the decay rate of the component over time and is extracted through the experimental data management system of the material supplier. represents the immediate contribution of various wear factors A i，j (t) to the wear state at the current time t. represents the aging trend of the component due to material properties and long-term fatigue effects.

[0034] In this embodiment, by constructing a wear prediction model based on the convolutional neural network (CNN) and combining historical time series response data with frequency domain analysis technology, intelligent modeling and dynamic prediction of the wear and aging trends of key components of industrial robots are realized. Not only are the historical vibration frequency characteristics fully extracted, but also by inputting the set A of response data groups in real time i for prediction calculation and outputting the aging wear prediction index wap, the aging degree of each component at each moment is effectively quantified. During the implementation process, by introducing the component material management system and the experimental data management system, the wear rate adjustment coefficient α i and the time decay coefficient γ i, realizing the differential modeling of the wear behaviors of different components. The aging wear prediction index wap comprehensively introduces real-time multi-source sensing data and material aging behaviors. By separately modeling the immediate wear effect and the time decay effect through a segmented structure, it realizes the dynamic health assessment of the key components of industrial robots under complex working conditions. Compared with traditional linear models, it is more physically reasonable and engineering adaptable. It can not only depict the decreasing influence trend of various wear factors over time, but also accurately simulate the irreversible aging process of materials accumulated with the operation cycle. Its structure is clear, the parameters are quantifiable, and it has good computability and generalization. It can significantly improve the accuracy of equipment fault warning and the scientific nature of maintenance strategies, and has outstanding practical value and innovation in industrial automation scenarios. This step significantly improves the ability of industrial robots to identify potential aging risks, transforms the operation and maintenance strategy from "fixed-cycle maintenance" to "condition prediction and regulation", reduces the sudden failure rate of equipment, and improves the continuity of the production line and the intelligent maintenance level.

[0035] Example 5

[0036] This example is an explanatory note carried out in Example 4. Please refer to Figure 1 , specifically: S4 includes S41 and S42; S41. After performing equipment aging prediction, the local server defines the functional dependencies and physical conduction relationships between the various components of the industrial robot based on the electronic structure drawings and component working link diagrams of the industrial robot, uses system modeling language and BOM structure list to identify the hierarchy and coupling relationships between equipment modules, and generates the aging influence propagation coefficient w using the historical collaborative fault data learning method ip , specifically as follows; ; w ip represents the causal intensity of the influence of the failure of component i on component p; S42. The local server performs comprehensive calculations based on the obtained health status score index hss, aging wear prediction index wap, and aging influence propagation coefficient w ip to obtain the aging response propagation index fpx and analyze the impact of the aging of a single component on the operation of the entire robot; ; In the formula, fpx p (t) represents the failure response propagation index of the aging component to the p-th component at time t, N represents the total number of robot components, e represents the exponential function, represents the inherent failure tendency of component i, that is, the potential probability of triggering a failure under the dual action of the current state and historical wear of the component, It represents the propagation ability and conduction efficiency of the risk of component i to the target component p, and analyzes the non-linear and multi-time scale propagation effects consistent with the actual system.

[0037] In this embodiment, using the electronic structure drawing of the industrial robot and the component working link diagram, combined with the system modeling language and the BOM structure list, the functional dependencies and coupling relationships between components are clarified, and the aging influence propagation coefficient w is generated based on learning historical collaborative fault data. ip Subsequently, the health state score index hss, the aging wear prediction index wap, and the aging influence propagation coefficient w are fused. ip The aging response propagation index fpx is comprehensively calculated, so as to accurately quantify the chain effect of the aging of any component on the overall system operation, realizing the dynamic quantification of the fault risk in the equipment system and the multi-source propagation path modeling. By introducing a non-linear time decay mechanism, it effectively reflects the linkage diffusion characteristics of faults evolving with time in a complex structure system. It improves the accuracy of hidden fault identification and prediction, and also realizes the priority early warning of key components and the closed-loop control of system-level risks, significantly enhancing the reliability of the equipment, the forward-looking of maintenance decisions, and the adaptive ability of the operating system. This implementation method not only establishes a dynamic feedback mechanism for the propagation of component aging to system-level faults, but also effectively overcomes the problem of the lack of structural conduction path modeling in traditional aging analysis methods, realizes the multi-dimensional mapping and deduction of aging risks from single points to systems, greatly improves the systematicness, forward-looking, and decision-making accuracy of fault early warning, and provides a controllable and adjustable optimization path for equipment maintenance and health management.

[0038] Embodiment 6

[0039] This embodiment is an explanatory description carried out in Embodiment 5. Please refer to Figure 1 and Figure 2 , specifically: S5 includes S51 and S52; S51. The local server non-weightedly accumulates the fault propagation probability values of all components of the industrial robot according to the obtained aging response propagation index fpx, calculates and obtains the comprehensive fault risk prediction index SRR, and analyzes the overall fault risk of the industrial robot. The specific formula is as follows; ; In the formula, \(\overline{f_{px}(t)}\) represents the mean value of the aging response propagation indices of all components at time t.

[0040] S52. Collect the comprehensive fault risk prediction index SRR of industrial robots in normal and faulty states at various historical time periods. Calculate the mean value of the historical comprehensive fault risk prediction index SRR according to the statistical method, and set a preset operation fault response threshold V based on the mean value. Then compare it with the obtained comprehensive fault risk prediction index SRR to conduct an evaluation of the comprehensive fault response of industrial robots. The specific evaluation scheme is as follows; When the comprehensive fault risk prediction index SRR ≥ the operation fault response threshold V, it indicates that there are aging components in the industrial robot, and the aging components affect the normal operation of other components. At this time, control the robot to stop immediately according to the robot control system, and generate component aging information and transmit it to the relevant personnel's user terminal through wireless communication to remind the replacement of aging equipment components; When the comprehensive fault risk prediction index SRR < the operation fault response threshold V, it indicates that the robot components are normal, and normal monitoring is maintained.

[0041] In this embodiment, after obtaining the aging response propagation index fpx, the comprehensive fault risk prediction index SRR is calculated by the non - weighted accumulation method, and the operation fault response threshold V is set in combination with the mean value of fpx in the historical operation state, realizing the quantitative risk assessment and accurate judgment of the overall operation state of industrial robots. Specifically, when the comprehensive fault risk prediction index SRR is greater than or equal to the operation fault response threshold V, a shutdown instruction is automatically generated and the component aging information is pushed, triggering the warning mechanism; if the comprehensive fault risk prediction index SRR is lower than the operation fault response threshold V, the device is kept in a normal operation state and continuously monitored. This method not only establishes a data - driven dynamic threshold evaluation mechanism, avoiding the problem of easy misjudgment of static alarms, but also realizes the efficient closed - loop management from component micro - aging to system macro - risk, significantly improving the timeliness of fault prevention, the intelligence of equipment management and the overall operation and maintenance efficiency, and finally achieving a comprehensive improvement effect of reducing equipment failure rate, extending service life and optimizing maintenance cycle.

[0042] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An AI-based pre-perception analysis method for the health status of IoT device terminals, characterized by: The following steps are involved: S1. Collecting the operation data of the industrial robot in real time according to the sensor group installed on the industrial robot, and transmitting it to the local server for preprocessing to obtain a response data group; S2. The local server performs collective processing on the response data group to form a response data group set A. i , and based on the response data set A i Calculate the health status score index HSS to conduct operational health assessment; S3. When the operation health assessment shows that the equipment is operating normally, a wear prediction model is constructed based on the convolutional neural network CNN, and then the acquired response data set A is obtained. i Input the wear prediction model to obtain the aging wear prediction index wap; S4. The local server defines the functional relationship of the components based on the electronic structure drawings and component work link diagrams of the industrial robot, and uses the historical collaborative fault data learning method to generate the aging impact propagation coefficient w ip , and then combine the health status score index hss and the aging wear prediction index wap to calculate the aging response propagation index fpx; S5. The local server performs non-weighted accumulation of the fault propagation probability values ​​of all components of the industrial robot based on the acquired aging response propagation index fpx, and calculates the comprehensive fault risk prediction index SRR for comprehensive fault response evaluation.

2. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 1 is characterized by: S1 includes S11 and S12; S11. Install sensor groups on key components of the industrial robot, and collect operating data of the industrial robot in real time through the installed sensor groups; The sensor group includes temperature sensor, piezoelectric sensor, vibration sensor, current sensor, voltage sensor and humidity sensor.

3. The AI-based IoT device terminal health status pre-perception analysis method according to claim 2 is characterized by: S12, transmitting the operating data collected in real time by the sensor group to a local server via a wireless network, and the local server extracting feature data related to the health status of the industrial robot from the operating data by using a principal component analysis technique; The principal component analysis technology extracts feature data related to the health status of industrial robots from multiple original features of the operating data by reducing the dimension; After obtaining the feature data, the local server performs data cleaning, denoising, outlier detection and dimensionless processing on the feature data to obtain a response data group; Data cleaning automatically removes outlier data through the set data quality threshold. De-noising uses the Kalman filter algorithm to weightedly fuse historical data with current feature data to filter out the noise impact in feature data. Outlier detection uses a machine learning model combined with historical data to identify abnormal patterns in operating data. Dimensionless processing uses the Max-Min method to eliminate the dimensional impact of operating data. The response data set includes operating temperature wd, operating load fz, vibration amplitude zf, current intensity dl, voltage intensity yq and humidity sd.

4. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 3 is characterized by: S2 includes S21; S21. The local server collects the response data set of the same i-th component of the industrial robot collected at time t to form a response data set set A. i , expressed as: A i ⊇{wd, fz, zf, dl, yq, sd}, and store it in the data repository after marking the timestamp.

5. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 4 is characterized in that: S2 also includes S22; S22, based on the acquired response data set A i Analyze the health status of each component of the industrial robot, and then analyze the health status of the industrial robot, and conduct a health assessment of the industrial robot, specifically including S221 and S222; S221, the local server is based on set A i Calculate and obtain the health status score index hss, the specific calculation formula is as follows; ； In the formula, hss i (t) represents the health status score index of the i-th component at time t, m represents the total number of collected parameters, k i，j A represents the adjustment factor of the jth parameter on the health status score index of the ith component, i，j (t) represents the jth parameter value collected in real time by the i-th component at time t, B i，j represents the standard value of the jth parameter of the ith component under standard working conditions, λ j represents the time decay factor of the jth parameter, exp represents the exponential decay function, t represents the current time t, and t0 represents the initial time when the i-th component starts working.

6. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 5 is characterized by: S222. Based on the healthy operation industry standard in the field of industrial robot equipment health management, a preset mechanical operation health threshold X is performed, and then compared with the obtained health status score index hss to perform an operation health assessment. The specific assessment plan is as follows; When the health status score index hss ≤ the mechanical operation health threshold X, it means that there is a fault in the equipment. At this time, the robot control system controls the machine to stop immediately, and generates fault information, which is transmitted to the relevant personnel user end through wireless communication to remind the machine fault that maintenance is required; When the health status score index hss> the mechanical operation health threshold X, it means that the equipment is operating normally, and the equipment prediction instruction is executed at this time.

7. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 6 is characterized by: S3 includes S31 and S32; S31. When the operation health assessment shows that the equipment is operating normally, the local server constructs a wear prediction model based on the convolutional neural network CNN and queries the time series response data set A of the historical industrial robot equipment in the data repository. i , analyze the time series characteristics of the historical response data group, then extract the frequency components in the equipment vibration signal through fast Fourier transform, analyze the changes of the historical response data group in different frequency ranges, obtain the historical time series characteristic response data set, input the historical time series characteristic response data set into the wear prediction model for model training, and optimize the wear prediction model; S32, the response data set A obtained in real time i The trained wear prediction model is input to predict equipment aging. The wear prediction model outputs the aging wear prediction index wap. The specific formula for analyzing the aging and wear of industrial robot equipment is as follows; ； In the formula, wap i (t) represents the aging wear prediction index of the i-th component at time t, m represents the total number of collected parameters, α i represents the adjustment coefficient of the wear rate of the i-th component, α i，0 represents the initial wear coefficient of the i-th component, e represents the exponential function, γ i represents the time decay coefficient of the ith component.

8. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 7 is characterized in that: S4 includes S41 and S42; S41. After executing the equipment aging prediction, the local server defines the functional dependency and physical conduction relationship between the various components of the industrial robot based on the electronic structure drawings and component work link diagrams of the industrial robot, uses the system modeling language and BOM structure list to identify the hierarchy and coupling relationship between equipment modules, and uses the historical collaborative fault data learning method to generate the aging impact propagation coefficient w ip , as follows; ； w ip It represents the causal strength of the impact on component p when component i fails; S42, the local server obtains the health status score index hss, the aging wear prediction index wap and the aging impact propagation coefficient w ip Perform comprehensive calculations to obtain the aging response propagation index fpx and analyze the impact of the aging of a single component on the operation of the entire robot; ； Where, fpx p (t) represents the fault response propagation index of the aging component to the pth component at time t, N represents the total number of robot components, and e represents the exponential function.

9. The AI-based pre-perception analysis method for the health status of IoT device terminals according to claim 8 is characterized by: S5 includes S51; S51. The local server performs non-weighted accumulation of the fault propagation probability values ​​of all components of the industrial robot according to the acquired aging response propagation index fpx, calculates and obtains the comprehensive fault risk prediction index SRR, and analyzes the overall fault risk of the industrial robot. The specific formula is as follows; ； In the formula, (t) represents the mean of the aging response propagation index of all components at time t.

10. The AI-based pre-perception analysis method for the health status of an IoT device terminal according to claim 9 is characterized in that: S5 also includes S52; S52. Collect the comprehensive fault risk prediction index SRR of normal and faulty industrial robots in each historical period, calculate the mean of the historical comprehensive fault risk prediction index SRR according to the statistical method, and preset the operation fault response threshold V based on the mean, and then compare it with the obtained comprehensive fault risk prediction index SRR to evaluate the comprehensive fault response of the industrial robot. The specific evaluation plan is as follows; When the comprehensive fault risk prediction index SRR ≥ the operating fault response threshold V, it means that the industrial robot has aging components, and the aging components affect the normal operation of other components. At this time, the robot control system controls the machine to stop immediately, and generates component aging information and transmits it to the relevant personnel user end through wireless communication to remind the equipment components to be replaced due to aging; When the comprehensive fault risk prediction index SRR is less than the operating fault response threshold V, it means that the robot components are normal and maintain normal monitoring.

Images