Hardware equipment intelligent monitoring method based on Internet of Things

Abstract

The invention discloses an intelligent hardware equipment monitoring method based on the Internet of Things, which comprises the following steps of: acquiring data such as sensor signals, temperature power consumption of a processor, voltage and current of a power supply module and the like in real time, processing the data by applying technologies such as a filtering algorithm, a causal inference model and time sequence analysis and judging the performance change of a component; and inputting the processed data into a pre-training regression model, predicting the residual life of each component and generating a health score, integrating the health scores of each component, introducing a dynamic correlation analysis and adaptive weight adjustment mechanism, constructing a system-level health state report, and generating a priority maintenance plan for high-risk components based on the report. Intelligent evaluation and prediction of the health state of the hardware component are realized, interruption of operation of the monitoring system due to hardware faults is effectively avoided, and the reliability and maintenance efficiency of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of device monitoring, and in particular to an intelligent monitoring method for hardware devices based on the Internet of Things. Background Art

[0002] With the rapid development of Internet of Things technology, intelligent monitoring systems have been widely applied in many fields. The Internet of Things intelligent monitoring system realizes real-time monitoring and data collection of the target environment or object through hardware components such as sensors, processors, and power modules. However, during the long-term operation of existing intelligent monitoring systems, potential problems of hardware devices have gradually emerged, becoming a key factor affecting the stability and reliability of the system.

[0003] Firstly, monitoring devices usually need to work continuously all day long. Hardware components such as sensors, processors, and power modules are prone to performance degradation or failure under high-load operating conditions. For example, when the sensor continuously collects data, its sensitivity may gradually decrease due to environmental temperature, humidity, or physical wear, resulting in a serious impact on the accuracy and timeliness of monitoring data. In addition, in high-concurrency data processing scenarios, the temperature rise and power consumption fluctuations of the processor may lead to a decline in chip performance, and even crashes or restarts, affecting the stable operation of the monitoring system.

[0004] Secondly, during the long-term power supply process of the power module, problems such as capacitor aging and unstable voltage may cause device power-off or damage, resulting in monitoring interruption. These potential problems usually have concealment and progression, and are difficult to identify in a timely manner through conventional monitoring means. Traditional fault detection methods often rely on after-sales maintenance and cannot give early warnings before problems occur, resulting in a relatively high fault risk for the monitoring system.

[0005] In addition, although there are some studies on device performance degradation and fault prediction in the prior art, most of them focus on specific fields or single components, lacking systematic analysis and prediction of the performance degradation of the overall hardware devices in the Internet of Things intelligent monitoring system. For example, some patents propose performance degradation prediction methods based on machine learning or data-driven, but these methods often require a large amount of historical data for training in practical applications and have limited processing capabilities for real-time data. Summary of the Invention

[0006] The present invention provides an intelligent monitoring method for hardware devices based on the Internet of Things to solve the above-mentioned existing technical problems.

[0007] The technical solution of the present invention is realized as follows:

[0008] An intelligent monitoring method for hardware devices based on the Internet of Things includes the following steps:

[0009] Obtain real-time data of hardware components during operation, including signal values collected by sensors, temperature and power consumption of the processor, and voltage and current parameters of the power supply module;

[0010] For the sensor signal values, remove noise interference, extract the change trend of the signal, and determine whether the sensor sensitivity is lower than the preset threshold;

[0011] According to the temperature and power consumption data of the processor, introduce a causal inference model, combine the real-time data and historical failure records of the processor, and analyze the root cause of performance degradation;

[0012] Through joint modeling, predict potential performance problems;

[0013] Extract the voltage and current parameters of the power supply module, and judge whether there is a trend of voltage instability or capacitor aging through time series analysis;

[0014] Input the change trend of sensor sensitivity, temperature and power consumption fluctuations of the processor, and voltage stability data of the power supply module into a pre-trained regression model to predict the remaining life of each component;

[0015] According to the remaining life prediction results, generate a health score for the hardware components. Components with a score lower than the preset threshold are marked as high risk and trigger a maintenance warning;

[0016] Integrate the health scores of all components, construct a system-level health status report, introduce a dynamic correlation analysis and adaptive weight adjustment mechanism, and realize the intelligent evaluation and dynamic update of the system-level health status.

[0017] Furthermore, the process of obtaining real-time data of hardware components during operation includes:

[0018] Obtain the signal values collected by the sensors, the temperature value, power consumption value of the processor, the voltage value, and current value of the power supply module;

[0019] Perform data preprocessing on the obtained sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values to remove noise and outliers;

[0020] Adopt a data normalization method to normalize the preprocessed sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values.

[0021] Furthermore, the process of determining whether the sensor sensitivity is lower than the preset threshold includes:

[0022] Use the sensor to obtain the signal value of the hardware component, and transmit the signal value to the calculator through the collector;

[0023] The calculator processes the signal value through a filtering algorithm to remove noise interference and obtains the denoised signal value; extracts the signal change trend based on the denoised signal value to obtain trend line data;

[0024] Compares the trend line data with a preset sensitivity threshold to judge the sensor sensitivity state. If the sensor sensitivity is lower than the preset threshold, starts the sensitivity calibration algorithm to adjust the sensor parameters;

[0025] Re - collects the signal value through the calibrated sensor parameters to obtain new signal data. Performs filtering processing on the new signal data and extracts the updated signal change trend;

[0026] Makes a secondary judgment on the updated trend line data and the preset threshold to determine whether the sensor sensitivity meets the requirements.

[0027] Furthermore, the process of analyzing the root cause of performance degradation includes:

[0028] Obtains the real - time temperature value and power consumption value of the processor to generate an initial data set;

[0029] Adopts a causal inference model to analyze the causality between the temperature value and the power consumption value to obtain a preliminary causal analysis result;

[0030] Obtains the historical failure records of the processor, combines the real - time temperature value and power consumption value, and corrects the preliminary causal analysis result to obtain the corrected causality;

[0031] Through a joint modeling method, correlates the corrected causality with the performance value of the processor to determine the root cause of performance degradation;

[0032] If there is a root cause of performance degradation, generates a prediction result of potential performance problems based on potentiality and predictability.

[0033] Furthermore, the process of predicting potential performance problems includes:

[0034] Adopts a machine learning algorithm to optimize the operating parameters of the processor according to the prediction result of potential performance problems to obtain optimized parameter values;

[0035] Through the optimized parameter values, re - collects the temperature value and power consumption value of the processor to judge whether the performance problem is solved;

[0036] Extracts the voltage value and current value of the power supply module to generate time - series data;

[0037] Adopts a time - series analysis method to calculate the standard deviation and trend slope of voltage fluctuations, judges whether the voltage fluctuations exceed the allowable range, and if they exceed the range, marks the power supply module as a potential fault.

[0038] Further, the process of determining whether there is a trend of voltage instability or capacitor aging includes:

[0039] Obtain the voltage value and current value of the power supply module to generate time series data;

[0040] Adopt time series analysis method to calculate the standard deviation and trend slope of voltage fluctuation. According to the preset range, judge whether the voltage fluctuation exceeds the allowable range. If the voltage fluctuation exceeds the range, combine the capacitor aging trend slope to judge the degree of capacitor aging;

[0041] Mark the power supply module as a potential fault according to the voltage fluctuation and the degree of capacitor aging. Use machine learning algorithms to classify the fault types of the power supply modules with potential faults;

[0042] Generate a maintenance priority list for the power supply module according to the fault type classification results. Obtain the sensitivity change amount of the sensor, the temperature value and power consumption fluctuation data of the processor, and the voltage stability data of the power supply module;

[0043] Input the obtained sensitivity change amount, temperature value, power consumption value, and voltage value into the pre-trained regression model to obtain the remaining life prediction values of each component.

[0044] Further, the process of predicting the remaining life of each component includes:

[0045] Input the preprocessed sensitivity change trend data, temperature value, power consumption fluctuation data, and voltage stability data into the pre-trained regression model;

[0046] Use the regression model to perform regression analysis on the input sensitivity change amount, temperature value, power consumption value, and voltage value to obtain the remaining life prediction values of each component.

[0047] Further, the process of triggering a maintenance warning includes:

[0048] Adopt a preset health score calculation formula, use the remaining life prediction value as an input variable, generate a health score value through the mapping relationship. If the health score value is lower than the preset threshold, start the high-risk marking mechanism and update the high-risk marking information in the component status database;

[0049] Trigger the maintenance warning mechanism according to the high-risk marking information to generate warning information including component identification, health score value, and remaining life value;

[0050] Use classification algorithms to analyze high-risk components to determine the fault types of the components. The fault types include performance degradation, function failure, and complete damage;

[0051] Calculate the maintenance priority of components based on the fault type and health score value. The maintenance priority is generated using a weighted algorithm that combines the fault type weight and the health score weight;

[0052] Use a sorting algorithm to sort all high-risk components based on the maintenance priority, generating a maintenance task queue;

[0053] Transmit the maintenance task queue to the maintenance management system, which schedules maintenance resources according to the task queue;

[0054] Adopt a dynamic priority algorithm to generate a maintenance priority list based on the health score and operating frequency of high-risk components;

[0055] Generate an initial maintenance plan according to the maintenance priority list, combined with the preset maintenance resource allocation rules.

[0056] Furthermore, the process of realizing the intelligent evaluation and dynamic update of the system-level health status includes:

[0057] Obtain the health score data of all components and calculate the system-level health score using a score integration algorithm;

[0058] Generate a system health status report based on the system-level health score, combined with the preset threshold range;

[0059] Extract the operation data of components from the system and identify the association rules between components using a dynamic association analysis algorithm;

[0060] Set an adaptive weight adjustment mechanism according to the association rules to dynamically adjust the health score weights of components;

[0061] Based on the adjusted health score weights, use an intelligent evaluation algorithm to calculate the comprehensive evaluation value of the system health status;

[0062] Judge whether the system health status has changed according to the comprehensive evaluation value. If it has changed, trigger the dynamic update mechanism;

[0063] Through the dynamic update mechanism, update the system health status report and transmit the updated report to the management system;

[0064] Extract the health score data of all components from the health status report, and mark the components with scores lower than the preset threshold as high-risk components;

[0065] Generate a maintenance priority list using a dynamic priority algorithm based on the health score and operating frequency of high-risk components.

[0066] Furthermore, it also includes: generating a maintenance plan according to the health status report, and giving priority to arranging maintenance or replacement for high-risk components;

[0067] The process of generating a maintenance plan includes:

[0068] Extract the health score data of all components from the health status report. Using a preset health score calculation formula, take the remaining life prediction value as an input variable, and generate a health score value through a mapping relationship. If the health score value is lower than the preset threshold, activate the high-risk marking mechanism and update the high-risk marking information in the component status database;

[0069] According to the high-risk marking information, trigger the maintenance warning mechanism and generate warning information including component identification, health score value, and remaining life value;

[0070] Use a classification algorithm to analyze high-risk components and determine the failure types of the components. The failure types include three types: performance degradation, function failure, and complete damage;

[0071] According to the failure type and health score value, use a weighted algorithm to combine the failure type weight and health score weight to calculate the maintenance priority of the component;

[0072] Use a sorting algorithm to sort all high-risk components. The sorting basis is the maintenance priority, and generate a maintenance task queue;

[0073] Through an association analysis algorithm, identify the dependency relationships between high-risk components and adjust the task order in the maintenance task queue;

[0074] Use a predictive maintenance model to estimate the remaining service life of high-risk components and optimize the task schedule in the maintenance task queue;

[0075] Transmit the optimized maintenance task queue to the maintenance management system. The maintenance management system schedules maintenance resources according to the task queue and generates maintenance task instructions.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] 1. By collecting various operation data of hardware components in real time and combining processing means such as denoising, trend analysis, and sensitivity calibration, the present invention ensures the accuracy and real-time nature of monitoring data. At the same time, it uses a causal inference model and machine learning algorithms to optimize the processor performance and solve the problem of performance degradation under high-load operation; and through time series analysis and failure classification algorithms, it warns of potential hidden dangers of the power supply module, and identifies and processes problems such as capacitor aging and voltage instability in advance, thereby effectively avoiding the performance attenuation and failure of hardware components;

[0078] 2. By generating maintenance warnings and maintenance plans, potential failures of hardware components are identified in advance. The remaining life of components is predicted using regression models, and combined with fault type classification and maintenance priority ranking, predictive maintenance is achieved. During the maintenance process, the system monitors the health status of high-risk components in real time and dynamically updates the maintenance plan to ensure that faults are handled before they occur, thus avoiding the drawbacks of traditional after-the-fact repairs and improving the stability and reliability of the system;

[0079] 3. By integrating the real-time data of sensors, processors, and power modules to form a comprehensive view, and using a joint modeling method to systematically analyze the performance degradation of the overall hardware device. Combining various technical means such as filtering algorithms, time series analysis, causal inference models, and machine learning algorithms, multi-dimensional processing and analysis of the performance data of hardware components are carried out to accurately identify the root causes of performance degradation; In addition, an optimized maintenance plan is generated through a dynamic priority algorithm and resource allocation rules, realizing comprehensive monitoring and predictive maintenance of the overall performance of the Internet of Things intelligent monitoring system, effectively making up for the deficiencies of existing technologies. Brief Description of the Drawings

[0080] Figure 1 It is a flowchart of the steps of an intelligent monitoring method for hardware devices based on the Internet of Things according to the present invention. Detailed Embodiments

[0081] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0082] As Figure 1 shown, an intelligent monitoring method for hardware devices based on the Internet of Things includes the following steps:

[0083] Obtain the real-time data of hardware components during operation, including the signal values collected by sensors, the temperature and power consumption of the processor, and the voltage and current parameters of the power module;

[0084] For the sensor signal values, remove noise interference, extract the change trend of the signal, and determine whether the sensor sensitivity is lower than the preset threshold;

[0085] According to the temperature and power consumption data of the processor, introduce a causal inference model, and combine the real-time data and historical fault records of the processor to analyze the root causes of performance degradation;

[0086] Predict potential performance issues through joint modeling;

[0087] Extract the voltage and current parameters of the power supply module, and determine whether there is a trend of unstable voltage or capacitor aging through time series analysis;

[0088] Input the sensitivity change trend of the sensor, the temperature and power consumption fluctuations of the processor, and the voltage stability data of the power supply module into the pre-trained regression model to predict the remaining life of each component;

[0089] Generate a health score for the hardware components based on the remaining life prediction results. Components with a score lower than the preset threshold are marked as high risk to trigger a maintenance warning;

[0090] Integrate the health scores of all components to build a system-level health status report, introduce a dynamic correlation analysis and adaptive weight adjustment mechanism, and realize the intelligent evaluation and dynamic update of the system-level health status.

[0091] Further, the process of obtaining the real-time data of the hardware components during operation includes:

[0092] Obtain the signal values collected by the sensor, the temperature value and power consumption value of the processor, and the voltage value and current value of the power supply module;

[0093] Preprocess the obtained sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values to remove noise and outliers;

[0094] Use data standardization methods to normalize the preprocessed sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values;

[0095] Input the normalized sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values into the pre-trained regression model, and analyze the sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values through the regression model to obtain the remaining life prediction values of each component;

[0096] Based on the remaining life prediction values of each component, determine whether the life value of each component is lower than the preset threshold. If the life value is lower than the preset threshold, determine that the component is the target component to be replaced or repaired;

[0097] Use a classification algorithm to classify the fault types of the target components to be replaced or repaired to obtain the fault type classification results of the target components;

[0098] Combine the remaining life prediction values and fault type classification results of the target components to determine the maintenance priority of each component;

[0099] Generate maintenance plans for each component according to the maintenance priority and update the system maintenance record;

[0100] In one embodiment, obtain the signal values collected by the sensors, the temperature value and power consumption value of the processor, and the voltage value and current value of the power supply module. For example, the sensor signal value is 0.5V, the processor temperature is 65°C, the power consumption value is 12W, the power supply module voltage is 5V, and the current is 2A;

[0101] Perform data preprocessing on the obtained sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values, remove noise and outliers, smooth the signal values using the moving average method, and use the 3σ criterion to eliminate outliers;

[0102] Adopt a data normalization method to normalize the preprocessed sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values, and use Z-score normalization to map the data to a distribution with a mean of 0 and a standard deviation of 1;

[0103] Input the normalized sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values into a pre-trained regression model, for example, use the random forest regression algorithm for modeling;

[0104] Analyze the sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values through the regression model to obtain the predicted remaining life values of each component. For example, the remaining life of the sensor is 1000 hours, the remaining life of the processor is 800 hours, and the remaining life of the power supply module is 1200 hours;

[0105] According to the predicted remaining life values of each component, judge whether the life values of each component are lower than the preset threshold. For example, the preset threshold is 500 hours. If the remaining life of the sensor is 400 hours, then determine that the component is the target component that needs to be replaced or repaired;

[0106] Adopt a classification algorithm to classify the fault types of the target components that need to be replaced or repaired. For example, use the support vector machine algorithm to classify the fault types into three categories: sensor sensitivity decline, processor overheating, and power supply module voltage instability;

[0107] Combine the predicted remaining life values of the target components and the fault type classification results to determine the maintenance priority of each component. For example, set the priority of processor overheating to the highest;

[0108] Generate maintenance plans for each component according to the maintenance priority and update the system maintenance record. For example, arrange for the processor to be replaced in the next maintenance cycle.

[0109] Further, the process of determining whether the sensor sensitivity is lower than a preset threshold includes:

[0110] Use a sensor to obtain the signal value of the hardware component, and transmit the signal value to a calculator through a collector;

[0111] The calculator performs a filtering algorithm on the signal value to remove noise interference and obtain a denoised signal value; extract the signal change trend based on the denoised signal value to obtain trend line data;

[0112] Compare the trend line data with the preset sensitivity threshold to determine the sensor sensitivity status. If the sensor sensitivity is lower than the preset threshold, start the sensitivity calibration algorithm and adjust the sensor parameters;

[0113] Re - collect the signal value through the calibrated sensor parameters to obtain new signal data. Perform filtering on the new signal data and extract the updated signal change trend;

[0114] Perform a secondary judgment on the updated trend line data and the preset threshold to determine whether the sensor sensitivity meets the requirements;

[0115] The display integrates all real - time data to form a comprehensive view of the operating state of the hardware component;

[0116] In one embodiment, use a sensor to obtain the signal value of the hardware component. For example, collect a voltage signal from a temperature sensor with a range of 0 - 5V, and transmit the signal value to a calculator through a collector.

[0117] The calculator performs a filtering algorithm on the signal value. For example, use Kalman filtering to remove noise interference, and control the fluctuation range of the filtered signal value within ±0.1V to obtain a denoised signal value;

[0118] Extract the signal change trend based on the denoised signal value. For example, generate a trend line through linear regression analysis with a regression coefficient R2 of 0.95 to obtain trend line data;

[0119] Compare the trend line data with the preset sensitivity threshold. For example, the threshold is 0.5V / ℃, and determine the sensor sensitivity status;

[0120] If the sensor sensitivity is lower than the preset threshold. For example, the current sensitivity is 0.4V / ℃, start the sensitivity calibration algorithm, use the least - squares method to adjust the sensor parameters, and increase the sensitivity to 0.6V / ℃. Re - collect the signal value through the calibrated sensor parameters. For example, collect 10 groups of data with an interval of 1 second between each group to obtain new signal data;

[0121] Perform filtering on the new signal data. For example, use median filtering to remove outliers and extract the updated signal change trend;

[0122] The updated trend line data is subjected to a secondary judgment with a preset threshold. For example, if the slope of the trend line is 0.55 V / °C, it is determined that the sensor sensitivity meets the requirements;

[0123] The display integrates all real-time data. For example, the sensor signal value, the processor temperature value, and the power module power value are displayed in the form of a chart to form a comprehensive view of the operating status of the hardware components.

[0124] Further, the process of analyzing the root cause of the performance degradation includes:

[0125] Obtain the real-time temperature value and power consumption value of the processor to generate an initial data set;

[0126] Adopt a causal inference model to analyze the causality between the temperature value and the power consumption value to obtain a preliminary causal analysis result;

[0127] Obtain the historical failure records of the processor, and combine the real-time temperature value and power consumption value to correct the preliminary causal analysis result to obtain the corrected causality;

[0128] Through a joint modeling method, associate the corrected causality with the performance value of the processor to determine the root cause of the performance degradation;

[0129] If there is a root cause of the performance degradation, generate a prediction result of potential performance problems according to potentiality and predictability.

[0130] Further, the process of predicting potential performance problems includes:

[0131] Adopt a machine learning algorithm to optimize the operating parameters of the processor according to the prediction result of potential performance problems to obtain optimized parameter values;

[0132] Based on the optimized parameter values, re-collect the temperature value and power consumption value of the processor to determine whether the performance problem is solved;

[0133] Extract the voltage value and current value of the power module to generate time series data;

[0134] Adopt a time series analysis method to calculate the standard deviation and trend slope of the voltage fluctuation, and determine whether the voltage fluctuation exceeds the allowable range. If it exceeds the range, mark the power module as a potential fault;

[0135] In one embodiment, obtain the real-time temperature value and power consumption value of the processor to generate an initial data set. For example, the temperature of the processor under load is collected as 75°C and the power consumption is 95 W;

[0136] Adopt a causal inference model, such as based on Pearson correlation coefficient and Granger causality test, to analyze the causality between temperature values and power consumption values, and obtain the preliminary causal analysis results. If the power consumption increases by 0.8W for every 1℃ increase in temperature value, it is determined that temperature has a significant causal impact on power consumption;

[0137] Obtain the historical failure records of the processor, such as performance degradation events caused by overheating in the past year, and combine the real-time temperature values and power consumption values to correct the preliminary causal analysis results. If the historical records show that the power consumption fluctuates more in high-temperature environments, then correct the causality to that the power consumption increases by 1.2W for every 1℃ increase in temperature;

[0138] Through a joint modeling method, such as constructing a multiple regression model, associate the corrected causality with the performance value of the processor to determine the root cause of performance degradation. If the increase in power consumption leads to a 5% performance degradation, and if there is a root cause of performance degradation, according to potentiality and predictability, such as using a Bayesian network to predict that the probability of performance degradation in the next week is 30%, generate the prediction results of potential performance problems;

[0139] Adopt a machine learning algorithm, such as an optimization algorithm based on decision trees, and optimize the operating parameters of the processor according to the prediction results of potential performance problems to obtain the optimized parameter values. For example, adjust the maximum power consumption limit to 90W;

[0140] With the optimized parameter values, re-collect the temperature values and power consumption values of the processor to determine whether the performance problem has been solved. For example, after optimization, the temperature stabilizes at 70℃ and the power consumption drops to 85W;

[0141] Extract the voltage value and current value of the power supply module to generate time series data. For example, collect the voltage value of 12V and the current value of 5A;

[0142] Adopt a time series analysis method, such as the ARIMA model, to calculate the standard deviation and trend slope of voltage fluctuations, and determine whether the voltage fluctuations exceed the allowable range. If the standard deviation is 0.5V and the trend slope is negative, mark the power supply module as potentially faulty.

[0143] Furthermore, the process of determining whether there is a trend of voltage instability or capacitor aging includes:

[0144] Obtain the voltage value and current value of the power supply module to generate time series data;

[0145] Adopt a time series analysis method to calculate the standard deviation and trend slope of voltage fluctuations. According to the preset range, determine whether the voltage fluctuations exceed the allowable range. If the voltage fluctuations exceed the range, combine the trend slope of capacitor aging to determine the degree of capacitor aging;

[0146] Mark the power supply module as a potential fault according to the voltage fluctuation and the degree of capacitor aging. Use a machine learning algorithm to classify the fault types of the potentially faulty power supply modules;

[0147] Generate a maintenance priority list for the power supply module according to the classification results of the fault types. Obtain the change amount of the sensor sensitivity, the temperature value and power consumption fluctuation data of the processor, and the voltage stability data of the power supply module;

[0148] Input the obtained change amount of sensitivity, temperature value, power consumption value, and voltage value into the pre-trained regression model to obtain the predicted remaining life values of each component;

[0149] In one embodiment, collect the voltage value and current value of the power supply module through the sensor, record and generate time series data at a frequency of once per second;

[0150] Use the moving average method and the autoregressive integrated moving average model (ARIMA) to analyze the voltage fluctuation, calculate its standard deviation as 0.05V, and the trend slope as -0.002V / hour;

[0151] According to the preset allowable range of voltage fluctuation of ±0.1V, determine whether the current voltage fluctuation exceeds the range. If the voltage fluctuation exceeds the range, combined with the capacitor aging trend slope of 0.0015V / hour, determine that the degree of capacitor aging is moderate;

[0152] Mark the power supply module as a potential fault according to the voltage fluctuation and the degree of capacitor aging.

[0153] Further, the process of predicting the remaining life of each component includes:

[0154] Input the preprocessed sensitivity change trend data, temperature value, power consumption fluctuation data, and voltage stability data into the pre-trained regression model;

[0155] Use the regression model to perform regression analysis on the input change amount of sensitivity, temperature value, power consumption value, and voltage value to obtain the predicted remaining life values of each component;

[0156] In one embodiment, collect the sensitivity change trend data of the sensor, and obtain its sensitivity change trend from the initial value of 1.0 to 0.8 through the sensor history record;

[0157] Obtain the temperature value of the processor, record the fluctuation data of its temperature rising from 25°C to 35°C in real time, and at the same time extract the range of the power consumption value fluctuating from 2.5W to 3.0W;

[0158] Extract the voltage stability data of the power supply module, and record the change of the voltage from 12.0V to 11.8V;

[0159] Preprocess the sensitivity change trend data, temperature values, power consumption fluctuation data, and voltage stability data. Use Gaussian filtering to remove noise and standardize the data to the same dimension through the Z-score normalization method;

[0160] Input the preprocessed sensitivity change of 0.2, temperature value of 35°C, power consumption value of 3.0W, and voltage value of 11.8V into the pre-trained random forest regression model;

[0161] Use the regression model to perform regression analysis on the input parameters, evaluate the model performance using the mean squared error, and obtain the predicted remaining life of the sensor as 5000 hours, the predicted remaining life of the processor as 3000 hours, and the predicted remaining life of the power supply module as 4000 hours.

[0162] Furthermore, the process of triggering the maintenance warning includes:

[0163] Adopt a preset health score calculation formula, use the predicted remaining life value as the input variable, generate a health score value through the mapping relationship. If the health score value is lower than the preset threshold, start the high-risk marking mechanism and update the high-risk marking information in the component status database;

[0164] According to the high-risk marking information, trigger the maintenance warning mechanism and generate warning information including component identification, health score value, and remaining life value;

[0165] Use a classification algorithm to analyze high-risk components and determine the failure types of the components. The failure types include three types: performance degradation, function failure, and complete damage;

[0166] According to the failure type and health score value, calculate the maintenance priority of the component. The maintenance priority is generated by combining the failure type weight and the health score weight using a weighted algorithm;

[0167] Use a sorting algorithm to sort all high-risk components. The sorting basis is the maintenance priority, and generate a maintenance task queue;

[0168] Transmit the maintenance task queue to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue;

[0169] Adopt a dynamic priority algorithm to generate a maintenance priority list according to the health score and operating frequency of high-risk components;

[0170] According to the maintenance priority list, combine the preset maintenance resource allocation rules to generate an initial maintenance plan;

[0171] In one embodiment, a preset health score calculation formula is adopted, with the remaining life prediction value as the input variable, and a health score value is generated through a mapping relationship. For example, a component with a remaining life of 1000 hours is calculated by the formula Score = (Lifetime / Max_Lifetime) * 100. If Max_Lifetime is 5000 hours, the score is 20. If the health score value is lower than the preset threshold of 30, a high-risk marking mechanism is activated, and the high-risk marking information is updated in the component status database. For example, the status field of component ID 1234 is marked as "high risk".

[0172] Based on the high-risk marking information, a maintenance warning mechanism is triggered to generate a warning message containing the component identification, health score value, and remaining life value. For example, the warning message is "Component 1234, health score 20, remaining life 1000 hours".

[0173] A classification algorithm is used to analyze high-risk components to determine the failure type of the components. For example, the KNN algorithm is used to classify component data, and the failure type is obtained as "performance degradation".

[0174] Based on the failure type and health score value, the maintenance priority of the component is calculated. The maintenance priority is generated by a weighted algorithm combining the failure type weight and the health score weight. For example, the failure type weight is 0.7, and the health score weight is 0.3, then the priority Score = 0.7 * Fault_Type + 0.3 * Health_Score. A sorting algorithm is used to sort all high-risk components, and the sorting basis is the maintenance priority, generating a maintenance task queue. For example, sorting from high to low according to the priority, generating the queue [component 1234, component 5678].

[0175] The maintenance task queue is transmitted to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue. For example, assign maintenance personnel A to handle component 1234.

[0176] A dynamic priority algorithm is adopted to generate a maintenance priority list according to the health score and operating frequency of high-risk components. For example, the priority of components with a high operating frequency is increased by 10%.

[0177] According to the maintenance priority list, combined with the preset maintenance resource allocation rules, an initial maintenance plan is generated. For example, give priority to arranging maintenance for components with a high operating frequency and a low health score.

[0178] Further, the process of realizing the intelligent evaluation and dynamic update of the system-level health status includes:

[0179] Obtain the health score data of all components, and use a score integration algorithm to calculate the system-level health score.

[0180] Generate a system health status report based on the system-level health score and in combination with a preset threshold range;

[0181] Extract the operation data of components from the system and use a dynamic association analysis algorithm to identify the association rules between components;

[0182] Set an adaptive weight adjustment mechanism according to the association rules to dynamically adjust the health score weights of components;

[0183] Based on the adjusted health score weights, use an intelligent evaluation algorithm to calculate the comprehensive evaluation value of the system health status;

[0184] Judge whether the system health status has changed according to the comprehensive evaluation value. If it has changed, trigger the dynamic update mechanism;

[0185] Update the system health status report through the dynamic update mechanism and transmit the updated report to the management system;

[0186] Extract the health score data of all components from the health status report. For components with scores lower than the preset threshold, mark them as high-risk components;

[0187] Generate a maintenance priority list according to the health scores and operation frequencies of high-risk components using a dynamic priority algorithm;

[0188] In one embodiment, obtain the health score data of all components, use a weighted average score integration algorithm to perform weighted summation of the scores of each component according to the preset weights, and obtain a system-level health score of 85 points;

[0189] Generate a system health status report based on the system-level health score and in combination with a preset threshold range (90 points is healthy, 70 - 90 points is sub-healthy, and lower than 70 points is unhealthy). The report shows that the current system is in a sub-healthy state;

[0190] Extract the operation data of components from the system and use the Apriori dynamic association analysis algorithm to identify the association rules between components. It is found that there is a strong correlation between the operation states of component A and component B. According to the association rules, set an adaptive weight adjustment mechanism to adjust the weight of component A from 0.3 to 0.4 and the weight of component B from 0.2 to 0.3, and dynamically adjust the health score weights of components;

[0191] Based on the adjusted health score weights, use a fuzzy comprehensive evaluation intelligent evaluation algorithm to calculate the comprehensive evaluation value of the system health status as 82 points;

[0192] Judge whether the system health status has changed according to the comprehensive evaluation value. If the evaluation value drops by more than 5 points compared with the previous time, trigger the dynamic update mechanism;

[0193] Update the system health status report through a dynamic update mechanism and transmit the updated report to the management system;

[0194] Extract the health score data of all components from the health status report. For components with a score lower than the preset threshold of 70 points, mark them as high-risk components. For example, the score of component C is 65 points;

[0195] According to the health score and operating frequency of the high-risk components, use a dynamic priority algorithm to calculate that the priority of component C is high, generate a maintenance priority list, and list component C as the priority maintenance object.

[0196] Furthermore, it also includes: generating a maintenance plan based on the health status report, and giving priority to arranging maintenance or replacement for high-risk components;

[0197] The process of generating the maintenance plan includes:

[0198] Extract the health score data of all components from the health status report, use a preset health score calculation formula, take the remaining life prediction value as the input variable, generate a health score value through the mapping relationship. If the health score value is lower than the preset threshold, start the high-risk marking mechanism and update the high-risk marking information in the component status database;

[0199] According to the high-risk marking information, trigger the maintenance warning mechanism and generate warning information including component identification, health score value, and remaining life value;

[0200] Use a classification algorithm to analyze high-risk components to determine the component failure types, which include three types: performance degradation, function failure, and complete damage;

[0201] According to the failure type and health score value, use a weighted algorithm combined with the failure type weight and health score weight to calculate the maintenance priority of the component;

[0202] Use a sorting algorithm to sort all high-risk components based on the maintenance priority to generate a maintenance task queue;

[0203] Through an association analysis algorithm, identify the dependency relationships between high-risk components and adjust the task order in the maintenance task queue;

[0204] Use a predictive maintenance model to estimate the remaining service life of high-risk components and optimize the task time arrangement in the maintenance task queue;

[0205] Transmit the optimized maintenance task queue to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue to generate maintenance task instructions;

[0206] In one embodiment, health score data of all components is extracted from the health status report. Using a preset health score calculation formula, for example, the formula: Health Score = 100×(Predicted Remaining Life / Designed Life), with the predicted remaining life as the input variable, a health score value is generated through the mapping relationship. If the health score value is lower than the preset threshold of 60, a high-risk marking mechanism is initiated, and the high-risk marking information is updated in the component status database. For example, the health score of component A is marked as high risk;

[0207] Based on the high-risk marking information, a maintenance warning mechanism is triggered to generate a warning message containing the component identifier, health score value, and remaining life value. For example, a warning message is generated: Component A has a health score of 55 and a predicted remaining life of 6 months;

[0208] A classification algorithm is used to analyze high-risk components. For example, the decision tree algorithm is used to determine the failure types of the components, and the failure types include three types: performance degradation, function failure, and complete damage;

[0209] Based on the failure type and health score value, a weighted algorithm is used to combine the failure type weight and health score weight. For example, the failure type weight is 0.6 and the health score weight is 0.4, to calculate the maintenance priority of the component;

[0210] A sorting algorithm is used to sort all high-risk components. For example, the quicksort algorithm is used, and the sorting basis is the maintenance priority, to generate a maintenance task queue;

[0211] Through an association analysis algorithm, the dependency relationships between high-risk components are identified. For example, the association rule mining algorithm is used to adjust the task order in the maintenance task queue;

[0212] A predictive maintenance model is used to estimate the remaining service life of high-risk components. For example, a time series prediction model is used to optimize the task time arrangement in the maintenance task queue. The optimized maintenance task queue is transmitted to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue to generate a maintenance task instruction and transmit it to the execution module of the monitoring system.

[0213] The above only lists some preferred embodiments of the present invention, but the present invention is not limited thereto, and many improvements and transformations can be made. As long as the improvements and transformations are made based on the basic principles of the present invention, they should be regarded as falling within the protection scope of the present invention.

Claims

1. An intelligent monitoring method for hardware devices based on the Internet of Things, characterized in that, It includes the following steps: Obtain real-time data of hardware components during operation, including signal values collected by sensors, temperature and power consumption of the processor, and voltage and current parameters of the power supply module; For the sensor signal values, remove noise interference, extract the change trend of the signal, and determine whether the sensor sensitivity is lower than the preset threshold; According to the temperature and power consumption data of the processor, introduce a causal inference model, combine the real-time data and historical fault records of the processor, and analyze the root cause of performance degradation; Predict potential performance problems through joint modeling; Extract the voltage and current parameters of the power supply module, and judge whether there is a trend of voltage instability or capacitor aging through time series analysis; Input the change trend of sensor sensitivity, temperature and power consumption fluctuations of the processor, and voltage stability data of the power supply module into a pre-trained regression model to predict the remaining life of each component; Generate a health score for the hardware components according to the remaining life prediction results. Components with a score lower than the preset threshold are marked as high risk and trigger a maintenance warning; Integrate the health scores of all components, construct a system-level health status report, introduce a dynamic correlation analysis and adaptive weight adjustment mechanism, and realize the intelligent evaluation and dynamic update of the system-level health status.

2. The intelligent monitoring method for hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of obtaining real-time data of hardware components during operation includes: Obtain the signal values collected by sensors, the temperature value and power consumption value of the processor, and the voltage value and current value of the power supply module; Perform data preprocessing on the obtained sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values to remove noise and outliers; Adopt a data normalization method to normalize the preprocessed sensor signal values, processor temperature values, processor power consumption values, power supply module voltage values, and power supply module current values.

3. The intelligent monitoring method for hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of determining whether the sensor sensitivity is lower than the preset threshold includes: Use a sensor to obtain the signal value of the hardware component, and transmit the signal value to a calculator through a collector; The calculator processes the signal value with a filtering algorithm to remove noise interference and obtain the denoised signal value; extract the signal change trend according to the denoised signal value to obtain trend line data; Compare the trend line data with the preset sensitivity threshold to judge the sensor sensitivity state. If the sensor sensitivity is lower than the preset threshold, start the sensitivity calibration algorithm to adjust the sensor parameters; Collect signal values again through the calibrated sensor parameters to obtain new signal data. Perform filtering processing on the new signal data and extract the updated signal change trend; Perform a secondary judgment on the updated trend line data and the preset threshold to determine whether the sensor sensitivity meets the requirements.

4. A method for intelligent monitoring of hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of analyzing the root cause of performance degradation includes: Obtain the real-time temperature value and power consumption value of the processor and generate an initial data set; Adopt a causal inference model to analyze the causality between the temperature value and the power consumption value to obtain a preliminary causal analysis result; Obtain the historical fault records of the processor, combine the real-time temperature value and power consumption value, and correct the preliminary causal analysis result to obtain the corrected causality; By means of a joint modeling method, the corrected causality is associated with the performance value of the processor to determine the root cause of performance degradation; If there is a root cause of performance degradation, according to potentiality and predictability, a prediction result of potential performance problems is generated.

5. The intelligent monitoring method for hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of predicting potential performance problems includes: Using a machine learning algorithm, according to the prediction result of potential performance problems, the operating parameters of the processor are optimized to obtain optimized parameter values; Based on the optimized parameter values, the temperature value and power consumption value of the processor are re-collected to determine whether the performance problem is solved; Extract the voltage value and current value of the power supply module to generate time series data; Adopt a time series analysis method to calculate the standard deviation and trend slope of voltage fluctuations, and determine whether the voltage fluctuations exceed the allowable range. If the range is exceeded, mark the power supply module as potentially faulty.

6. The intelligent monitoring method for hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of determining whether there is a trend of voltage instability or capacitor aging includes: Obtain the voltage value and current value of the power supply module to generate time series data; Adopt a time series analysis method to calculate the standard deviation and trend slope of voltage fluctuations. According to the preset range, determine whether the voltage fluctuations exceed the allowable range. If the voltage fluctuations exceed the range, combined with the trend slope of capacitor aging, determine the degree of capacitor aging; Based on the voltage fluctuations and the degree of capacitor aging, mark the power supply module as potentially faulty. Use a machine learning algorithm to classify the fault types of the potentially faulty power supply module; According to the fault type classification result, generate a maintenance priority list for the power supply module. Obtain the sensitivity change amount of the sensor, the temperature value and power consumption fluctuation data of the processor, and the voltage stability data of the power supply module; Input the obtained sensitivity change amount, temperature value, power consumption value, and voltage value into a pre-trained regression model to obtain the predicted remaining life values of each component.

7. A method for intelligent monitoring of hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of predicting the remaining life of each component includes: Input the pre-processed sensitivity change trend data, temperature value, power consumption fluctuation data, and voltage stability data into a pre-trained regression model; Use the regression model to perform regression analysis on the input sensitivity change amount, temperature value, power consumption value, and voltage value to obtain the predicted remaining life values of each component.

8. The intelligent monitoring method for hardware devices based on the Internet of Things according to claim 1, wherein, The process of triggering a maintenance warning includes: Using a preset health score calculation formula, taking the predicted remaining life value as an input variable, generating a health score value through a mapping relationship. If the health score value is lower than the preset threshold, start a high-risk marking mechanism and update the high-risk marking information in the component status database; According to the high-risk marking information, trigger a maintenance warning mechanism to generate warning information including component identification, health score value, and remaining life value; Use a classification algorithm to analyze high-risk components to determine the fault types of the components. The fault types include three types: performance degradation, function failure, and complete damage; According to the fault type and health score value, calculate the maintenance priority of the component. The maintenance priority is generated by a weighted algorithm combining the fault type weight and the health score weight; Use a sorting algorithm to sort all high-risk components. The sorting basis is the maintenance priority to generate a maintenance task queue; Transfer the maintenance task queue to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue; Adopt a dynamic priority algorithm to generate a maintenance priority list based on the health scores and operating frequencies of high-risk components; Generate an initial maintenance plan according to the maintenance priority list, combined with the preset maintenance resource allocation rules.

9. A method for intelligent monitoring of hardware devices based on the Internet of Things according to claim 1, characterized in that, The process of realizing the intelligent evaluation and dynamic update of the system-level health status includes: Obtain the health score data of all components, and use a score integration algorithm to calculate the system-level health score; Generate a system health status report according to the system-level health score, combined with the preset threshold range; Extract the operation data of components from the system, and use a dynamic correlation analysis algorithm to identify the correlation rules between components; Set an adaptive weight adjustment mechanism according to the correlation rules to dynamically adjust the health score weights of components; Based on the adjusted health score weights, use an intelligent evaluation algorithm to calculate the comprehensive evaluation value of the system health status; Judge whether the system health status has changed according to the comprehensive evaluation value. If it has changed, trigger the dynamic update mechanism; Through the dynamic update mechanism, update the system health status report and transmit the updated report to the management system; Extract the health score data of all components from the health status report, and mark the components with scores lower than the preset threshold as high-risk components; Generate a maintenance priority list according to the health scores and operating frequencies of high-risk components, using a dynamic priority algorithm.

10. A method for intelligent monitoring of hardware devices based on the Internet of Things according to claim 1, characterized in that, It also includes: Generate a maintenance plan according to the health status report, and give priority to arranging maintenance or replacement for high-risk components; The process of generating the maintenance plan includes: Extract the health score data of all components from the health status report, use a preset health score calculation formula, take the remaining life prediction value as the input variable, generate a health score value through the mapping relationship. If the health score value is lower than the preset threshold, start the high-risk marking mechanism and update the high-risk marking information in the component status database; Trigger the maintenance warning mechanism according to the high-risk marking information, and generate a warning message including component identification, health score value and remaining life value; Use a classification algorithm to analyze high-risk components to determine the failure types of components. The failure types include three types: performance degradation, function failure and complete damage; According to the failure type and health score value, use a weighted algorithm to combine the failure type weight and health score weight to calculate the maintenance priority of the component; Use a sorting algorithm to sort all high-risk components. The sorting basis is the maintenance priority, and generate a maintenance task queue; Identify the dependency relationships between high-risk components through the correlation analysis algorithm, and adjust the task order in the maintenance task queue; Adopt a predictive maintenance model to estimate the remaining service life of high-risk components and optimize the task time arrangement in the maintenance task queue; Transfer the optimized maintenance task queue to the maintenance management system, and the maintenance management system schedules maintenance resources according to the task queue to generate a maintenance task instruction.

Images