A computer hardware predictive maintenance early warning device and method based on running state perception

Abstract

The application discloses a computer hardware predictive maintenance early warning device and method based on running state sensing, and threshold self-adaptive creep calibration is realized by relying on an early warning calibration module in combination with hardware degradation entropy and equipment coupling auxiliary parameters, the threshold can be smoothly and dynamically adjusted with hardware aging and running environment, and the defects of threshold solidification, poor working condition adaptation and frequent false and missed reports are solved; secondly, a working condition modeling module fuses real-time parameters and hardware historical data to construct a working condition time sequence atlas, relies on multi-dimensional time sequence characteristics to analyze hardware states, can distinguish temporary load fluctuation from continuous hardware degradation, and improves degradation identification precision; meanwhile, a degradation evaluation module matches and compares real-time parameters with the atlas, calculates a unified quantitative hardware degradation entropy value to represent a decay degree, provides a quantitative basis for threshold dynamic adjustment and risk classification early warning, and solves the problems that traditional monitoring cannot quantify overall hardware degradation and fault risk identification is insufficient.

Description

Technical Field

[0001] This invention relates to the field of computer hardware maintenance and early warning technology, and in particular to a predictive maintenance and early warning device and method for computer hardware based on operational status perception. Background Technology

[0002] With the continuous improvement of computer hardware integration, the core hardware of equipment such as servers, industrial control hosts, and office terminals, such as CPUs, memory, hard drives, and motherboards, are in a state of continuous high load operation for a long time. Hardware aging, performance degradation, and hidden faults have the characteristics of gradual development. Sudden downtime and hardware damage can easily cause business interruption and data loss. Therefore, predictive maintenance and early warning of hardware has become a core requirement in the field of equipment operation and maintenance.

[0003] The current mainstream computer hardware status monitoring solutions mostly adopt a fixed threshold alarm mechanism: the upper and lower limits of single operating parameters such as temperature, occupancy rate, and read / write rate are set in advance. When the collected parameters exceed the threshold range, an alarm is triggered directly. However, this existing technology has obvious defects: (1) The threshold is static and fixed, and the adaptability is poor: the normal operating range of hardware varies greatly under different hardware models, different business loads, and different usage environments. The fixed threshold set by the user cannot adapt to different operating conditions. Low-load devices are prone to frequent false alarms, while high-load devices will miss early degradation signals due to the loose threshold. At the same time, the hardware continues to age with the years of use, and the tolerance range shifts year by year. The static threshold cannot be dynamically adjusted with the degree of hardware aging, and the timeliness of the warning continues to decline. (2) It only relies on instantaneous parameters for judgment and lacks multi-dimensional operating condition modeling capabilities: the existing monitoring methods only make simple numerical comparisons of parameters collected at a single time point. They do not integrate and model multi-dimensional information such as long-term historical operating data of hardware, load fluctuations, environmental interference, and life cycle loss. They cannot build complete hardware time-series operating condition characteristics, and it is difficult to accurately distinguish between instantaneous load fluctuations and continuous hardware degradation. The degradation identification accuracy is low. Summary of the Invention

[0004] In view of this, the present invention proposes a predictive maintenance and early warning device and method for computer hardware based on operational status perception, which can effectively solve the defects of existing technologies such as static and fixed thresholds, poor adaptability, and lack of multi-dimensional working condition modeling capabilities.

[0005] The technical solution of this invention is implemented as follows:

[0006] A predictive maintenance and early warning device for computer hardware based on operational status awareness, comprising:

[0007] The status acquisition module is used to collect the operating parameters of the computer's core hardware in real time.

[0008] The operating condition modeling module is used to receive real-time operating parameter datasets and combine them with historical hardware operating data to construct operating condition time series graphs.

[0009] The degradation assessment module is used to match and compare the real-time operating parameter dataset with the operating condition time series map to calculate the hardware degradation entropy value, which characterizes the degree of hardware operation degradation.

[0010] The early warning calibration module is used to perform threshold adaptive creep calibration on preset fault early warning thresholds based on hardware degradation entropy values ​​and combined with equipment operation coupling auxiliary parameters.

[0011] The terminal early warning module is used to receive hardware degradation entropy values, compare them with the calibrated fault early warning thresholds, and output hardware predictive maintenance early warning information in a tiered manner.

[0012] As a further optional solution to the aforementioned computer hardware predictive maintenance and early warning device based on operational status awareness, the operational condition modeling module includes:

[0013] The temporal registration and screening unit is used to perform temporal heterogeneous parameter registration on historical operating data and real-time operating parameter datasets to generate hardware state primitive samples.

[0014] Multidimensional topology partitioning unit is used to spatially mesh hardware state primitive samples with load disturbance, environmental stress and lifetime loss as three-dimensional coupling axes to generate temporal topology nodes carrying state offset weights.

[0015] The graph edge building and encapsulation unit is used to establish bidirectional evolutionary association edges of time-series topology nodes, encapsulate the hardware normal reference domain and degradation transition boundary corresponding to each node, and obtain the initial operating condition time-series graph.

[0016] The self-consistent evolutionary iterative unit is used to access incremental running samples in real time and dynamically update the offset weights and bidirectional evolutionary association edges of the time-series topology nodes in the initial operating condition time-series graph.

[0017] As a further optional solution to the aforementioned computer hardware predictive maintenance and early warning device based on operational status awareness, the degradation assessment module includes:

[0018] The state offset solving unit is used to match the real-time running parameter dataset with the hardware normal reference domain and the degradation transition boundary of the time-series topology node, and calculate the hardware multidimensional state offset.

[0019] The multi-dimensional weighted fusion unit is used to couple and weight the multi-dimensional state offset of hardware based on three types of weights: load disturbance, environmental stress, and life cycle loss, to generate an unordered feature set of hardware state.

[0020] The degradation entropy value derivation unit is used to obtain the hardware degradation entropy value, which characterizes the degree of hardware degradation, based on probability distribution fitting and information entropy algorithm, combined with the disordered feature set of hardware state.

[0021] As a further optional solution to the aforementioned computer hardware predictive maintenance and early warning device based on operational status awareness, the degradation entropy value inference unit includes:

[0022] The normalization constraint subunit is used to perform normalization mapping on the disordered feature set of hardware state to generate an disordered sample distribution sequence.

[0023] The boundary intercept correction subunit is used to retrieve the hardware normal reference domain and degradation transition boundary corresponding to the time-series topology node, and perform boundary intercept filtering on the disordered sample distribution sequence to obtain effective degradation feature samples.

[0024] The probability distribution fitting subunit is used to introduce multi-dimensional state coupling probability weights, perform temporal density fitting on effective degradation feature samples, and construct a disordered probability model of hardware operating state that adapts to the dynamic operating characteristics of hardware.

[0025] The global entropy value solving subunit is used to perform global statistical solutions on the disordered probability model of hardware operating state based on the information entropy operator, and output the hardware degradation entropy value that characterizes the degree of hardware operation degradation.

[0026] As a further optional solution to the aforementioned computer hardware predictive maintenance early warning device based on operational status awareness, the early warning calibration module includes:

[0027] Gradient vector construction unit is used to construct hardware stage-specific degradation gradient feature vectors based on hardware degradation entropy values ​​and device operation coupling auxiliary parameters.

[0028] The benchmark threshold matching unit is used to search the historical maintenance threshold adaptation sample library based on the time-series topology node attributes of the current operating condition time-series graph, match and retrieve the initial benchmark warning threshold range corresponding to the current hardware operating condition.

[0029] The progressive creep correction unit is used to progressively correct the initial baseline warning threshold range based on the gradient magnitude in the hardware stage degradation gradient feature vector as a dynamic correction coefficient.

[0030] The boundary fault tolerance compensation unit is used to combine the residual environmental stress and life cycle loss in the equipment operation coupling auxiliary parameters to perform boundary fault tolerance compensation on the threshold range after progressive creep correction, and output the final calibrated fault warning threshold.

[0031] As a further optional embodiment of the aforementioned predictive maintenance and early warning device for computer hardware based on operational status awareness, the gradient vector construction unit includes:

[0032] The timing window extraction subunit is used to extract timing data of hardware degradation entropy values ​​of multiple consecutive frames within a preset sliding timing window, and simultaneously extract the device operation coupling auxiliary parameters matched by the window time period;

[0033] The gradient magnitude calculation sub-unit is used to perform piecewise linear fitting on the hardware degradation entropy value within the window, solve the temporal change slope corresponding to the entropy value of each dimension, and generate the degradation gradient magnitude of each dimension.

[0034] The vector coupling assembly subunit is used to perform dimensional binding mapping between the magnitude of the degradation gradient in each dimension and the corresponding device operation coupling auxiliary parameters, so as to obtain the multi-component coupled hardware stage degradation gradient feature vector.

[0035] As a further optional embodiment of the aforementioned computer hardware predictive maintenance early warning device based on operational status awareness, the progressive creep correction unit includes:

[0036] The gradient magnitude extraction subunit is used to receive the hardware stage degradation gradient feature vector, parse and extract the degradation gradient magnitude of each dimension in the hardware stage degradation gradient feature vector as a dynamic correction coefficient.

[0037] Offset step size mapping sub-unit is used to construct a positive mapping rule between gradient magnitude and single threshold creep offset step size;

[0038] The smooth iterative creep sub-unit is used to iteratively adjust the upper and lower boundaries of the initial baseline warning threshold interval according to the single threshold creep offset step size obtained by mapping, so as to complete the gradual creep correction.

[0039] As a further optional embodiment of the aforementioned computer hardware predictive maintenance early warning device based on operational status awareness, the terminal early warning module includes:

[0040] The degradation level determination unit is used to perform a multi-dimensional comparison between the hardware degradation entropy value and the calibrated fault warning threshold, calculate the deviation of the hardware degradation entropy value relative to the fault warning threshold, and locate the current degradation level of the hardware.

[0041] The hierarchical strategy matching unit is used to match the corresponding maintenance and early warning handling strategy based on the current hardware degradation level.

[0042] A predictive maintenance and early warning method for computer hardware based on operational status awareness, specifically including:

[0043] Real-time acquisition of operating parameters of various core computer hardware components;

[0044] Based on real-time operating parameter datasets and combined with historical hardware operating data, a time series graph of operating conditions is constructed.

[0045] Based on the matching and comparison of real-time operating parameter datasets and operating condition time series graphs, the hardware degradation entropy value, which characterizes the degree of hardware operation degradation, is calculated.

[0046] Based on the hardware degradation entropy value and combined with the equipment operation coupling auxiliary parameters, the preset fault warning threshold is subjected to threshold adaptive creep calibration.

[0047] Based on the hardware degradation entropy value and compared with the calibrated fault warning threshold, predictive maintenance warning information for hardware is output in a tiered manner.

[0048] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described predictive maintenance and early warning method for computer hardware based on operational status awareness.

[0049] The beneficial effects of this invention are as follows: First, by combining the hardware degradation entropy value with the equipment operation coupling auxiliary parameters through the early warning calibration module, adaptive creep calibration of the early warning threshold is implemented, abandoning the fixed static threshold mode. The threshold can be dynamically and smoothly adjusted with hardware aging and operating environment, solving the defects of existing technologies such as static and fixed thresholds, poor adaptability to different hardware operating conditions, and easy generation of false alarms and missed alarms. Second, by integrating real-time operating parameters and historical hardware operating data through the operating condition modeling module to construct an operating condition time series map, hardware status analysis is completed based on multi-dimensional time series operating condition characteristics, making up for the shortcomings of existing technologies that rely only on instantaneous parameter judgment and lack multi-dimensional operating condition modeling capabilities. It can effectively distinguish between temporary load fluctuations and continuous hardware degradation, improving the accuracy of hardware degradation identification. At the same time, the degradation assessment module matches and compares real-time parameters with the operating condition time series map to calculate a unified and quantified hardware degradation entropy value to characterize the degree of hardware decay, providing a standardized quantitative basis for dynamic threshold adjustment and risk classification judgment, solving the problems of traditional monitoring methods being unable to quantify the overall degree of hardware degradation and having low fault risk identification. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a schematic diagram of the composition of a computer hardware predictive maintenance early warning device based on operational status perception according to the present invention.

[0052] Figure 2 This is a flowchart illustrating a predictive maintenance and early warning method for computer hardware based on operational status awareness, according to the present invention.

[0053] Figure 3 This is a schematic diagram of the composition of a computing device according to the present invention. Detailed Implementation

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] refer to Figures 1 to 3 A predictive maintenance and early warning device for computer hardware based on operational status awareness includes a status acquisition module, an operational condition modeling module, a degradation assessment module, an early warning calibration module, and a terminal early warning module, wherein:

[0056] The status acquisition module is used to collect the operating parameters of the computer's core hardware in real time.

[0057] Specifically, the status acquisition module is deployed at the computer's underlying driver level, interfacing with all core hardware sensor interfaces such as the CPU, memory, hard drive, and motherboard power supply chip, to continuously collect hardware operating parameters.

[0058] The data collection targets cover various core computer hardware such as servers, industrial control hosts, and office terminals. The collected parameters include real-time hardware occupancy rate, operating temperature, read / write speed, power supply voltage, continuous online running time, and ambient temperature and humidity.

[0059] Set a fixed acquisition cycle of 2 seconds to continuously collect real-time operating parameters of various hardware devices in batches, and package them into a standardized real-time operating parameter dataset. The dataset includes a collection timestamp, hardware device number, and device scenario identifier to ensure the traceability of time-series data.

[0060] After the data collection is completed, the complete parameter dataset is pushed to the backend working condition modeling module in real time, while the original collected data for 7 days is cached locally to prevent data loss due to transmission interruption.

[0061] A working condition modeling module is used to receive real-time operating parameter datasets and construct a working condition time series graph by combining them with historical hardware operating data; in some embodiments, the working condition modeling module includes:

[0062] The temporal registration and screening unit is used to perform temporal heterogeneous parameter registration on historical operating data and real-time operating parameter datasets to generate hardware state primitive samples.

[0063] Multidimensional topology partitioning unit is used to spatially mesh hardware state primitive samples with load disturbance, environmental stress and lifetime loss as three-dimensional coupling axes to generate temporal topology nodes carrying state offset weights.

[0064] The graph edge building and encapsulation unit is used to establish bidirectional evolutionary association edges of time-series topology nodes, encapsulate the hardware normal reference domain and degradation transition boundary corresponding to each node, and obtain the initial operating condition time-series graph.

[0065] The self-consistent evolutionary iterative unit is used to access incremental running samples in real time and dynamically update the offset weights and bidirectional evolutionary association edges of the time-series topology nodes in the initial operating condition time-series graph.

[0066] Specifically, the time-series registration and screening unit:

[0067] Historical operational data and real-time operational parameter datasets are read. The two types of data have different collection frequencies and parameter dimensions, and belong to heterogeneous parameters. Based on a unified 2-second time series scale, linear interpolation is performed to complete the missing historical data, and mean denoising is performed to screen redundant real-time collected data. The two types of data are uniformly mapped to the same time series scale, and abnormal outlier samples are removed. The hardware status primitive samples with time series alignment and unified dimensions are output.

[0068] Multidimensional topological subdivision unit:

[0069] We set load disturbance, environmental stress, and lifetime loss as three-dimensional coupled coordinate axes: load disturbance corresponds to CPU / memory usage and hard disk read / write load;

[0070] Environmental stress parameters correspond to computer room temperature and humidity, and dust concentration;

[0071] Lifecycle wear and tear corresponds to the cumulative runtime of the hardware and the power-on aging time.

[0072] The hardware state primitive samples are mapped onto the three-dimensional coordinate axis, and spatial gridding is performed to generate independent temporal topology nodes for each grid. The state offset weight is assigned according to the degree to which the sample deviates from the standard operating range of the hardware. The higher the offset weight, the stronger the hardware degradation trend of the node.

[0073] Graph edge building encapsulation unit:

[0074] Bidirectional evolutionary association edges are constructed between adjacent time-series topology nodes based on chronological order. Forward association edges represent the state evolution path of hardware as it naturally ages over time, while reverse association edges are used to trace back historical operating conditions. For each time-series topology node, all normal samples within the grid are statistically analyzed to define the hardware normal baseline domain, and samples with performance degradation or fault precursors are statistically analyzed to define the degradation transition boundary. The baseline domain, transition boundary, state offset weight, and bidirectional association edges are uniformly encapsulated to output the initial operating condition time-series graph.

[0075] Self-consistent evolutionary iterative unit:

[0076] It continuously receives incremental real-time running samples from the status acquisition module, matches the new samples to the corresponding time-series topology nodes of the 3D mesh, updates the state offset weight values ​​of the nodes according to the incremental samples, and synchronously adjusts the association strength of the bidirectional evolutionary association edges. It completes a lightweight iterative update of the graph every hour, realizing dynamic self-adaptation of the working condition time-series graph to follow the long-term operating status of the hardware, without the need for manual remodeling.

[0077] In this way, the multi-dimensional topology partitioning unit divides hardware operating conditions into three dimensions: load, environment, and life cycle, fully covering all core factors affecting hardware aging. It constructs time-series topology nodes with state offset weights, which, compared with traditional single-dimensional data comparison methods, has complete multi-dimensional operating condition modeling capabilities and can accurately distinguish between short-term load fluctuations and continuous hardware degradation. Secondly, the graph edge building and encapsulation unit connects all time-series operating conditions through bidirectional evolutionary related edges, while encapsulating the normal baseline domain and degradation transition boundary of each node, providing a standardized comparison reference for the subsequent degradation assessment module, and greatly improving the accuracy of hardware degradation degree calculation.

[0078] A degradation assessment module is used to match and compare real-time operating parameter datasets with operating condition time-series graphs to calculate a hardware degradation entropy value characterizing the degree of hardware performance degradation; in some embodiments, the degradation assessment module includes:

[0079] The state offset solving unit is used to match the real-time running parameter dataset with the hardware normal reference domain and the degradation transition boundary of the time-series topology node, and calculate the hardware multidimensional state offset.

[0080] The multi-dimensional weighted fusion unit is used to couple and weight the multi-dimensional state offset of hardware based on three types of weights: load disturbance, environmental stress, and life cycle loss, to generate an unordered feature set of hardware state.

[0081] The degradation entropy value derivation unit is used to obtain the hardware degradation entropy value, which characterizes the degree of hardware degradation, based on probability distribution fitting and information entropy algorithm, combined with the disordered feature set of hardware state.

[0082] Specifically, the state offset solution unit:

[0083] The current real-time operating parameter dataset is mapped to the three-dimensional coupled space of the operating condition time series graph, and the time series topology node corresponding to the parameter is matched. The hardware normal baseline domain and degradation transition boundary built into the node are retrieved, and the difference between each operating parameter such as CPU load, hard disk read and write, equipment temperature, and service life and the normal baseline domain is calculated. Then, the interval constraint correction is performed in combination with the degradation transition boundary to eliminate the extreme value interference caused by short-term load shocks. The hardware multi-dimensional state offset is output. The larger the offset value, the more significant the deviation of the current hardware operating state from the normal range.

[0084] Multidimensional weighted fusion unit:

[0085] The three-dimensional coupling axes of the working condition time series map are extracted to form three types of dynamic weights: load disturbance, environmental stress, and life cycle loss. The weight values ​​are updated synchronously with the offset weights of the topology nodes. The hardware state offsets of each dimension obtained in the previous step are multiplied by the corresponding dimension weights to complete the coupling weighting operation, and then normalized to the standard range of 0 to 1. Invalid small fluctuation components with amplitudes below the threshold after weighting are removed, and a standardized hardware state disorder feature set is assembled to fully integrate multi-dimensional hardware deviation information.

[0086] Degradation entropy value derivation unit:

[0087] The system receives an unordered feature set of hardware states. First, it performs interval normalization mapping on the multi-dimensional heterogeneous parameters of the feature set, constraining all parameters to a unified probability interval to generate an unordered sample distribution sequence. Then, it retrieves the degradation transition boundary of the matching time-series topology node to complete the boundary intercept correction and filter out meaningless steady-state fluctuations in the baseline domain. Subsequently, it introduces state coupling probability weights to fit the time-series probability distribution density and constructs an unordered probability model of hardware operating states. Finally, it uses the information entropy calculation formula to solve for the entropy value across the entire domain and outputs a hardware degradation entropy value in the range of 0 to 1. The closer the value is to 1, the higher the overall degree of hardware degradation. The stable hardware degradation entropy value is pushed to the early warning calibration module for threshold calibration.

[0088] Thus, the state offset calculation unit relies on the topological nodes of the operating condition time-series graph, the normal baseline domain, and the degradation transition boundary constructed above to complete the offset calculation. Based on a complete multi-dimensional historical operating condition as a reference, rather than a single instantaneous parameter comparison, it can effectively distinguish between temporary business load fluctuations and continuous hardware performance degradation, solving the shortcomings of existing technologies that lack multi-dimensional operating condition modeling and have insufficient degradation identification accuracy. Secondly, the multi-dimensional weighted fusion unit combines the weights of three dimensions—load, environment, and life cycle—to couple all offsets, integrating multi-dimensional hardware degradation information to generate a unified feature set, avoiding the one-sidedness caused by single-parameter judgment. In addition, the degradation entropy value inference unit outputs a unified quantified hardware degradation entropy value through probability fitting and information entropy algorithm, forming a continuously comparable hardware attenuation quantification index, making up for the shortcomings of traditional monitoring methods that can only judge whether parameters exceed the standard and cannot quantify the overall degree of hardware aging. The quantified entropy value can serve as the core calculation basis for the early warning calibration module, providing standardized input for threshold adaptive creep calibration, indirectly improving the problems of static threshold fixation, poor adaptability, and frequent false alarms and missed alarms in existing technologies.

[0089] In some embodiments, the degradation entropy value deduction unit includes:

[0090] The normalization constraint subunit is used to perform normalization mapping on the disordered feature set of hardware state to generate an disordered sample distribution sequence.

[0091] The boundary intercept correction subunit is used to retrieve the hardware normal reference domain and degradation transition boundary corresponding to the time-series topology node, and perform boundary intercept filtering on the disordered sample distribution sequence to obtain effective degradation feature samples.

[0092] The probability distribution fitting subunit is used to introduce multi-dimensional state coupling probability weights, perform temporal density fitting on effective degradation feature samples, and construct a disordered probability model of hardware operating state that adapts to the dynamic operating characteristics of hardware.

[0093] The global entropy value solving subunit is used to perform global statistical solutions on the disordered probability model of hardware operating state based on the information entropy operator, and output the hardware degradation entropy value that characterizes the degree of hardware operation degradation.

[0094] Specifically, normalized constraint sub-units:

[0095] The unordered feature set of hardware status includes multi-dimensional values ​​such as CPU load offset, hard disk wear offset, and environmental temperature and humidity offset. The dimensions and numerical ranges of each parameter vary greatly. This unit adopts the min-max normalization algorithm to uniformly map all heterogeneous offset parameters to the standard probability range of 0~1, eliminates the computational interference caused by the difference in magnitude of different parameters, and regularizes the output of a continuous and ordered unordered sample distribution sequence.

[0096] Boundary intercept correction sub-unit:

[0097] Retrieve the hardware normal baseline domain and degradation transition boundary stored in the current matching time-series topology node in the operating condition time-series graph; traverse all feature values ​​in the disordered sample distribution sequence, directly filter out small fluctuation values ​​that fall within the normal baseline domain range and represent normal hardware operation, and retain only feature data that exceeds the normal baseline domain and touches the degradation transition range, thus obtaining effective degradation feature samples after removing invalid steady-state interference.

[0098] Probability distribution fitting subunit:

[0099] Multidimensional state coupling probability weights corresponding to load disturbance, environmental stress, and life cycle loss are introduced, and the weight values ​​are consistent with the three-dimensional coupling axis weights set in the multidimensional topology partitioning unit. Using a 30-frame sliding time window as a unit, Gaussian time density fitting is performed on the effective degradation feature samples to generate a disordered probability model of hardware operating state that fits the long-term dynamic load and aging law of server hardware. The model can intuitively reflect the probability distribution of the occurrence of various degradation states of hardware.

[0100] Sub-unit for global entropy calculation:

[0101] The standard information entropy operator is called to perform global statistical calculations on the disordered probability model of hardware operating state. The higher the degree of dispersion of the probability distribution, the larger the entropy value obtained in the final solution. The calculation result is constrained to the interval of 0~1 to output the hardware degradation entropy value. The closer the entropy value is to 1, the higher the overall hardware operation degradation and the higher the risk of failure. The hardware degradation entropy value is continuously pushed to the early warning calibration module as the core input data for gradient vector construction.

[0102] Thus, the boundary intercept correction subunit reuses the normal baseline domain and degradation transition boundary of the topology nodes of the operating condition time series graph, filters out invalid fluctuation data in the normal operating range of hardware, retains only the true degradation characteristics, accurately distinguishes between instantaneous load disturbances and continuous hardware aging, further enhances the identification advantages brought by multi-dimensional operating condition modeling, and alleviates the defects of existing technology in degradation identification that are prone to misjudgment; secondly, the probability distribution fitting subunit integrates the probability weights of the three-dimensional coupling dimension to construct a dedicated probability model, which fully takes into account the three types of influencing factors: load, environment, and hardware life cycle. Compared with the simple single-parameter judgment method, it can comprehensively reflect the overall degradation trend of hardware, and the quantification results are more reliable.

[0103] An early warning calibration module is used to perform threshold adaptive creep calibration on a preset fault early warning threshold based on hardware degradation entropy values ​​and coupled auxiliary parameters of equipment operation; in some embodiments, the early warning calibration module includes:

[0104] Gradient vector construction unit is used to construct hardware stage-specific degradation gradient feature vectors based on hardware degradation entropy values ​​and device operation coupling auxiliary parameters.

[0105] The benchmark threshold matching unit is used to search the historical maintenance threshold adaptation sample library based on the time-series topology node attributes of the current operating condition time-series graph, match and retrieve the initial benchmark warning threshold range corresponding to the current hardware operating condition.

[0106] The progressive creep correction unit is used to progressively correct the initial baseline warning threshold range based on the gradient magnitude in the hardware stage degradation gradient feature vector as a dynamic correction coefficient.

[0107] The boundary fault tolerance compensation unit is used to combine the residual environmental stress and life cycle loss in the equipment operation coupling auxiliary parameters to perform boundary fault tolerance compensation on the threshold range after progressive creep correction, and output the final calibrated fault warning threshold.

[0108] Specifically, the gradient vector building unit:

[0109] A 30-frame sliding timing window is set to capture the hardware degradation entropy values ​​output within a continuous time period, and the corresponding device operation coupling auxiliary parameters are synchronously bound to the window. Piecewise linear fitting is performed on the temporal data of entropy values ​​within the window to solve the slope of entropy value changes corresponding to the three dimensions of load, environment, and life cycle as the degradation gradient magnitude of each dimension. The three types of gradient magnitudes are bound and encapsulated with the corresponding coupling auxiliary parameter dimensions to generate a multi-component coupled hardware stage degradation gradient feature vector, which fully represents the degradation evolution rate of hardware in different dimensions.

[0110] Benchmark threshold matching unit:

[0111] The system reads the time-series topology node attributes of the operating condition time-series graph matched with the current real-time parameters and retrieves the historical maintenance threshold adaptation sample library stored locally. The sample library pre-stores historical normal and fault operating condition samples and matching benchmark threshold ranges corresponding to each topology node. The system completes the retrieval and matching based on the node's load disturbance, environmental stress, and life cycle loss attributes, and retrieves the initial benchmark warning threshold range that is adapted to the current hardware operating condition. This range is the uncorrected static basic threshold.

[0112] Progressive creep correction unit:

[0113] The gradient magnitudes of each dimension within the feature vector of the hardware stage degradation gradient are extracted as dynamic correction coefficients. A positive mapping rule between gradient magnitude and threshold single offset step size is established: the larger the gradient magnitude, the faster the hardware degradation accelerates, and the corresponding threshold single creep offset step size is synchronously amplified. The upper and lower boundaries of the initial baseline warning threshold are finely adjusted one by one with a very small fixed iteration step size. The threshold boundary is constrained to have no instantaneous jumps throughout the process, completing a smooth and uninterrupted gradual creep correction and avoiding frequent alarm switching caused by threshold abrupt changes.

[0114] Boundary fault-tolerant compensation unit:

[0115] Two indicators, residual environmental stress and life cycle loss, are extracted from the auxiliary parameters coupled with equipment operation. The environmental compensation coefficient and aging compensation coefficient are calculated respectively. The two sets of compensation coefficients are superimposed on the upper and lower boundaries of the threshold range after creep correction. The threshold boundary is widened to compensate for the fault tolerance, which offsets the deterioration identification deviation caused by high temperature and humidity and long-term aging. The final calibrated dynamic fault warning threshold is output and pushed to the terminal warning module.

[0116] Thus, the gradient vector construction unit obtains the degradation gradient feature vector by fitting the continuous hardware degradation entropy value with the time-series sliding window. This can quantify the rate of hardware degradation and provide a quantitative correction basis for dynamic threshold adjustment. It connects the complete link of front-end multi-dimensional operating condition modeling and entropy value quantification evaluation, avoiding the one-sided correction logic caused by relying solely on a single point of entropy value for threshold adjustment. Secondly, the benchmark threshold matching unit reuses the time-series topology node attributes of the operating condition time-series map to match the initial benchmark threshold. The benchmark threshold conforms to the operating rules of similar hardware operating conditions in history. Compared with a uniformly preset fixed threshold, it improves the adaptability of the threshold to the current hardware operating condition from the source, alleviating the defects of poor adaptability and easy false alarms of existing static thresholds. In addition, the progressive creep correction unit dynamically controls the threshold smooth offset amplitude with the degradation gradient amplitude and adopts a progressive fine-tuning method to achieve adaptive creep adjustment of the threshold. There will be no alarm logic break caused by threshold jumps. The threshold can dynamically change with the real-time degradation rate of the hardware, solving the pain point that traditional fixed thresholds cannot adapt to the aging state of the entire hardware life cycle.

[0117] In some embodiments, the gradient vector construction unit includes:

[0118] The timing window extraction subunit is used to extract timing data of hardware degradation entropy values ​​of multiple consecutive frames within a preset sliding timing window, and simultaneously extract the device operation coupling auxiliary parameters matched by the window time period;

[0119] The gradient magnitude calculation sub-unit is used to perform piecewise linear fitting on the hardware degradation entropy value within the window, solve the temporal change slope corresponding to the entropy value of each dimension, and generate the degradation gradient magnitude of each dimension.

[0120] The vector coupling assembly subunit is used to perform dimensional binding mapping between the magnitude of the degradation gradient in each dimension and the corresponding device operation coupling auxiliary parameters, so as to obtain the multi-component coupled hardware stage degradation gradient feature vector.

[0121] Specifically, the timing window extracts sub-units:

[0122] The preset sliding timing window length is 30 frames. Combined with the data acquisition cycle of 2 seconds / time of the status acquisition module, a single window covers 60 seconds of continuous hardware degradation entropy value timing data. The window continuously captures multiple frames of hardware degradation entropy values ​​in real time. Based on the timestamp attached to each frame of entropy value, the corresponding device operation coupling auxiliary parameters for the same time period are synchronously matched and extracted. The entropy value timing data and the matching coupling auxiliary parameters are packaged into a window sample set. Abnormal window samples with missing acquisitions or value jumps are removed. The normalized timing samples are output to the downstream gradient magnitude solving sub-unit.

[0123] Gradient magnitude solution sub-unit:

[0124] Piecewise linear fitting is performed on the hardware degradation entropy value within the window based on three dimensions: load disturbance, environmental stress, and life cycle loss. The linear slope of the entropy value changing with time under the three dimensions is solved respectively. The slope values ​​are normalized and used as the degradation gradient magnitude of the corresponding dimension. A positive gradient magnitude indicates that the degree of hardware degradation continues to worsen. The larger the absolute value of the magnitude, the faster the hardware degradation evolution rate. The degradation gradient magnitudes of the three independent dimensions are output to the vector coupling assembly subunit.

[0125] Vector-coupled assembly subunits:

[0126] A three-dimensional one-to-one binding mapping relationship is established: the gradient magnitude of the load dimension is bound to the average daily load, the gradient magnitude of the environment dimension is bound to the temperature and humidity of the data center, and the gradient magnitude of the life cycle dimension is bound to the cumulative power-on time of the hardware. The three sets of gradient magnitudes and corresponding coupled auxiliary parameters are encapsulated into vector components to complete the dimension binding mapping. Finally, a hardware stage degradation gradient feature vector containing three types of coupled components of load, environment, and life cycle is generated, which fully conveys the correlation information between the hardware degradation rate and external operating conditions and is pushed to the progressive creep correction unit as the basis for calculating the dynamic correction coefficient.

[0127] Thus, the time-series window extraction subunit relies on the continuously output hardware degradation entropy values ​​from the front end and the synchronously acquired coupled auxiliary parameters to extract continuous time-series data using a fixed sliding window. This eliminates the single-point instantaneous entropy value judgment mode, enabling the capture of long-term hardware degradation trends and connecting the complete data link between front-end multi-dimensional operating condition modeling and entropy value quantification assessment. Secondly, the gradient magnitude calculation subunit solves for the slope of entropy value changes in each dimension through piecewise linear fitting, quantifying the rate of hardware degradation evolution in different dimensions. This intuitively distinguishes between slow aging and accelerated degradation conditions, providing quantifiable data for subsequent threshold-based creep correction. The dynamic adjustment basis addresses the shortcomings of traditional static thresholds, which cannot identify the rate of hardware degradation and have a single adjustment logic. In addition, the vector coupling assembly subunit binds the gradient magnitude with the device operation coupling auxiliary parameters to assemble a feature vector, and simultaneously integrates multiple information such as hardware degradation rate, environment, and life cycle, avoiding the one-sided adjustment caused by relying solely on the degradation rate as a single indicator for threshold correction. This vector is directly supplied to the progressive creep correction unit to complete the smooth fine adjustment of the threshold, supporting the implementation of the threshold adaptive creep calibration mechanism, and fundamentally improving the pain points of existing technology such as static and fixed thresholds, poor adaptability to different operating conditions, and frequent false alarms and missed alarms.

[0128] In some embodiments, the progressive creep correction unit includes:

[0129] The gradient magnitude extraction subunit is used to receive the hardware stage degradation gradient feature vector, parse and extract the degradation gradient magnitude of each dimension in the hardware stage degradation gradient feature vector as a dynamic correction coefficient.

[0130] Offset step size mapping sub-unit is used to construct a positive mapping rule between gradient magnitude and single threshold creep offset step size;

[0131] The smooth iterative creep sub-unit is used to iteratively adjust the upper and lower boundaries of the initial baseline warning threshold interval according to the single threshold creep offset step size obtained by mapping, so as to complete the gradual creep correction.

[0132] Specifically, the gradient magnitude extraction subunit:

[0133] The system receives the hardware stage degradation gradient feature vector output by the vector coupling assembly subunit, and performs analytical decomposition on the three components of load disturbance, environmental stress, and lifetime loss within the vector. The degradation gradient magnitude stored in each component is extracted as an independent dynamic correction coefficient. The three types of gradient magnitudes are weighted and fused to obtain a comprehensive correction coefficient. The value of the comprehensive correction coefficient directly reflects the overall hardware degradation acceleration. The comprehensive correction coefficient is output to the offset step size mapping subunit.

[0134] Offset step size mapping sub-unit:

[0135] The preset positive linear mapping rule is as follows: the larger the comprehensive correction coefficient (gradient magnitude), the larger the single threshold creep offset step size is. The mapping levels are divided according to the numerical range of the server's degradation entropy value from 0 to 1. Gradient magnitudes of 0 to 0.2 correspond to small offset step sizes, 0.2 to 0.6 correspond to medium offset step sizes, and 0.6 to 1 correspond to large offset step sizes. The extracted comprehensive correction coefficient is substituted into the mapping rule to match and output the single creep offset step size corresponding to this threshold adjustment, avoiding the problem that manually fixed step sizes cannot adapt to the differences in degradation rate.

[0136] Smooth iterative creep subunit:

[0137] The upper and lower boundaries of the initial baseline warning threshold range are adjusted step by step with a very small fixed number of iterations. Each round only adds the single offset step size obtained by mapping, and the complete threshold correction is completed in 10 iterations. The single adjustment range of the upper and lower boundaries of the threshold is constrained to not exceed the preset upper limit throughout the process to prevent the threshold boundary from changing abruptly. After the iteration is completed, the threshold range with progressive smooth creep correction is output and sent to the boundary fault tolerance compensation unit to carry out secondary fault tolerance compensation.

[0138] Thus, the offset step size mapping subunit establishes a positive correlation rule between gradient magnitude and threshold offset step size. The faster the hardware degradation rate, the larger the single threshold adjustment magnitude, which can adapt to the threshold adjustment needs under different hardware aging rates and solve the defects of traditional static thresholds that are fixed and cannot dynamically adapt to the degradation development. Secondly, the smooth iterative creep subunit uses a multi-round micro-iteration method to adjust the threshold range, realizing smooth creep correction without jumps, avoiding frequent alarm state switching and a large number of invalid interference alarms caused by threshold abrupt changes. The threshold range after smooth correction is then optimized by the boundary fault tolerance compensation unit, and finally outputs a dynamic warning threshold that adapts to the current hardware's full-dimensional operating conditions. This addresses the core pain points of existing technologies, such as static threshold fixation, poor operating condition adaptability, and frequent false alarms and missed alarms.

[0139] A terminal early warning module is used to receive hardware degradation entropy values, compare them with the calibration-completed fault early warning thresholds, and output tiered predictive maintenance early warning information for hardware; in some embodiments, the terminal early warning module includes:

[0140] The degradation level determination unit is used to perform a multi-dimensional comparison between the hardware degradation entropy value and the calibrated fault warning threshold, calculate the deviation of the hardware degradation entropy value relative to the fault warning threshold, and locate the current degradation level of the hardware.

[0141] The hierarchical strategy matching unit is used to match the corresponding maintenance and early warning handling strategy based on the current hardware degradation level.

[0142] Specifically, the degradation level determination unit:

[0143] The system obtains the dynamic fault warning threshold after calibration, and divides the dynamic fault warning threshold into two levels: a primary routine threshold and a secondary inspection threshold. It then performs a multi-dimensional numerical comparison between the real-time hardware degradation entropy value and the two thresholds, calculating the deviation of the entropy value from the threshold boundaries of each level. Based on the magnitude of the deviation, it classifies the system into three degradation levels: entropy values ​​below the primary threshold indicate mild degradation, entropy values ​​between the primary and secondary thresholds indicate moderate degradation, and entropy values ​​above the secondary threshold indicate severe degradation. The judgment process reuses the output results of the multi-dimensional operating condition modeling and adaptive threshold calibration described earlier, relying on dynamic thresholds to complete accurate level division and avoid the level judgment distortion caused by fixed thresholds.

[0144] Hierarchical strategy matching unit:

[0145] Pre-set maintenance early warning and handling strategies corresponding to three levels of degradation:

[0146] Mild degradation level: Matches the routine maintenance reminder strategy and generates monthly planned inspection reminder messages;

[0147] Moderate degradation level: Match the key inspection and early warning strategy to generate a special hardware detection alarm message within 48 hours;

[0148] Severe degradation level: Match emergency fault handling strategy and generate an emergency alarm message for immediate shutdown and maintenance; The message is synchronously bound with the operating condition data of the currently matched time-series topology node, the auxiliary parameters of equipment operation coupling, the hardware degradation entropy value, and the degradation gradient feature vector source information, and retains the full-link analysis basis. The corresponding level of early warning message is pushed to the data center operation and maintenance terminal for display in real time.

[0149] Thus, the degradation level determination unit completes multi-dimensional entropy comparison based on the dynamic threshold after front-end adaptive creep calibration. It uses dynamic thresholds that adapt to hardware aging, environmental changes, and load variations to classify the degradation levels. Compared to traditional fixed threshold grading methods, this accurately distinguishes between mild, moderate, and severe hardware degradation states, addressing the shortcomings of existing technologies such as static threshold fixation, poor operating condition adaptability, and ambiguous degradation risk identification. Secondly, the grading strategy matching unit matches differentiated maintenance and repair strategies based on the accurately determined degradation level. Supported by front-end multi-dimensional operating condition modeling, quantitative entropy value assessment, and dynamic threshold calibration, it achieves risk-layered early warning. Maintenance personnel can prioritize maintenance based on the warning level, eliminating the need for indiscriminate comprehensive inspections. This proactively avoids sudden hardware failures, ensures continuous and stable operation of data center servers, and significantly reduces the manpower costs associated with ineffective inspections, fully integrating the entire technical chain from data collection, operating condition modeling, degradation quantification, dynamic threshold calibration to grading warning output.

[0150] A predictive maintenance and early warning method for computer hardware based on operational status awareness, specifically including:

[0151] Real-time acquisition of operating parameters of various core computer hardware components;

[0152] Based on real-time operating parameter datasets and combined with historical hardware operating data, a time series graph of operating conditions is constructed.

[0153] Based on the matching and comparison of real-time operating parameter datasets and operating condition time series graphs, the hardware degradation entropy value, which characterizes the degree of hardware operation degradation, is calculated.

[0154] Based on the hardware degradation entropy value and combined with the equipment operation coupling auxiliary parameters, the preset fault warning threshold is subjected to threshold adaptive creep calibration.

[0155] Based on the hardware degradation entropy value and compared with the calibrated fault warning threshold, predictive maintenance warning information for hardware is output in a tiered manner.

[0156] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described predictive maintenance and early warning method for computer hardware based on operational status awareness.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A predictive maintenance and early warning device for computer hardware based on operational status awareness, characterized in that, include: The status acquisition module is used to collect the operating parameters of the computer's core hardware in real time. The operating condition modeling module is used to receive real-time operating parameter datasets and combine them with historical hardware operating data to construct operating condition time series graphs. The degradation assessment module is used to match and compare the real-time operating parameter dataset with the operating condition time series map to calculate the hardware degradation entropy value, which characterizes the degree of hardware operation degradation. The early warning calibration module is used to perform threshold adaptive creep calibration on preset fault early warning thresholds based on hardware degradation entropy values ​​and combined with equipment operation coupling auxiliary parameters. The terminal early warning module is used to receive hardware degradation entropy values, compare them with the calibrated fault early warning thresholds, and output hardware predictive maintenance early warning information in a tiered manner.

2. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 1, characterized in that, The working condition modeling module includes: The temporal registration and screening unit is used to perform temporal heterogeneous parameter registration on historical operating data and real-time operating parameter datasets to generate hardware state primitive samples. Multidimensional topology partitioning unit is used to spatially mesh hardware state primitive samples with load disturbance, environmental stress and lifetime loss as three-dimensional coupling axes to generate temporal topology nodes carrying state offset weights. The graph edge building and encapsulation unit is used to establish bidirectional evolutionary association edges of time-series topology nodes, encapsulate the hardware normal reference domain and degradation transition boundary corresponding to each node, and obtain the initial operating condition time-series graph. The self-consistent evolutionary iterative unit is used to access incremental running samples in real time and dynamically update the offset weights and bidirectional evolutionary association edges of the time-series topology nodes in the initial operating condition time-series graph.

3. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 2, characterized in that, The degradation assessment module includes: The state offset solving unit is used to match the real-time running parameter dataset with the hardware normal reference domain and the degradation transition boundary of the time-series topology node, and calculate the hardware multidimensional state offset. The multi-dimensional weighted fusion unit is used to couple and weight the multi-dimensional state offset of hardware based on three types of weights: load disturbance, environmental stress, and life cycle loss, to generate an unordered feature set of hardware state. The degradation entropy value derivation unit is used to obtain the hardware degradation entropy value, which characterizes the degree of hardware degradation, based on probability distribution fitting and information entropy algorithm, combined with the disordered feature set of hardware state.

4. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 3, characterized in that, The degradation entropy value deduction unit includes: The normalization constraint subunit is used to perform normalization mapping on the disordered feature set of hardware state to generate an disordered sample distribution sequence. The boundary intercept correction subunit is used to retrieve the hardware normal reference domain and degradation transition boundary corresponding to the time-series topology node, and perform boundary intercept filtering on the disordered sample distribution sequence to obtain effective degradation feature samples. The probability distribution fitting subunit is used to introduce multi-dimensional state coupling probability weights, perform temporal density fitting on effective degradation feature samples, and construct a hardware operating state disorder probability model that adapts to the dynamic operating characteristics of hardware. The global entropy value solving subunit is used to perform global statistical solutions on the disordered probability model of hardware operating state based on the information entropy operator, and output the hardware degradation entropy value that characterizes the degree of hardware operation degradation.

5. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 1, characterized in that, The early warning calibration module includes: Gradient vector construction unit is used to construct hardware stage-specific degradation gradient feature vectors based on hardware degradation entropy values ​​and device operation coupling auxiliary parameters. The benchmark threshold matching unit is used to search the historical maintenance threshold adaptation sample library based on the time-series topology node attributes of the current operating condition time-series graph, match and retrieve the initial benchmark warning threshold range corresponding to the current hardware operating condition. The progressive creep correction unit is used to progressively correct the initial baseline warning threshold range based on the gradient magnitude in the hardware stage degradation gradient feature vector as a dynamic correction coefficient. The boundary fault tolerance compensation unit is used to combine the residual environmental stress and life cycle loss in the equipment operation coupling auxiliary parameters to perform boundary fault tolerance compensation on the threshold range after progressive creep correction, and output the final calibrated fault warning threshold.

6. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 5, characterized in that, The gradient vector construction unit includes: The timing window extraction subunit is used to extract timing data of hardware degradation entropy values ​​of multiple consecutive frames within a preset sliding timing window, and simultaneously extract the device operation coupling auxiliary parameters matched by the window time period; The gradient magnitude calculation sub-unit is used to perform piecewise linear fitting on the hardware degradation entropy value within the window, solve the temporal change slope corresponding to the entropy value of each dimension, and generate the degradation gradient magnitude of each dimension. The vector coupling assembly subunit is used to perform dimensional binding mapping between the magnitude of the degradation gradient in each dimension and the corresponding device operation coupling auxiliary parameters, so as to obtain the multi-component coupled hardware stage degradation gradient feature vector.

7. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 5, characterized in that, The progressive creep correction unit includes: The gradient magnitude extraction subunit is used to receive the hardware stage degradation gradient feature vector, parse and extract the degradation gradient magnitude of each dimension in the hardware stage degradation gradient feature vector as a dynamic correction coefficient. Offset step size mapping sub-unit is used to construct a positive mapping rule between gradient magnitude and single threshold creep offset step size; The smooth iterative creep sub-unit is used to iteratively adjust the upper and lower boundaries of the initial baseline warning threshold interval according to the single threshold creep offset step size obtained by mapping, so as to complete the gradual creep correction.

8. The predictive maintenance and early warning device for computer hardware based on operational status awareness according to claim 1, characterized in that, The terminal early warning module includes: The degradation level determination unit is used to compare the hardware degradation entropy value with the calibrated fault warning threshold in multiple dimensions, calculate the deviation of the hardware degradation entropy value from the fault warning threshold, and locate the current degradation level of the hardware. The hierarchical strategy matching unit is used to match the corresponding maintenance and early warning handling strategy based on the current hardware degradation level.

9. A predictive maintenance and early warning method for computer hardware based on operational status awareness, characterized in that, Specifically, it includes: Real-time acquisition of operating parameters of various core computer hardware components; Based on real-time operating parameter datasets and combined with historical hardware operating data, a time series graph of operating conditions is constructed. Based on the matching and comparison of real-time operating parameter datasets and operating condition time series graphs, the hardware degradation entropy value, which characterizes the degree of hardware operation degradation, is calculated. Based on the hardware degradation entropy value and combined with the equipment operation coupling auxiliary parameters, the preset fault warning threshold is subjected to threshold adaptive creep calibration. Based on the hardware degradation entropy value and compared with the calibrated fault warning threshold, predictive maintenance warning information for hardware is output in a tiered manner.

10. A computing device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the computer hardware predictive maintenance and early warning method based on operational status awareness as described in claim 9.

Images