Remote monitoring method and system for aviation obstruction light

Abstract

The present invention relates to the technical field of monitoring, in particular to a remote monitoring method and system for an aviation obstruction light. The method comprises the following steps: on the basis of an external sensor, acquiring electromagnetic signals sent by an aviation obstruction light; and by means of using a signal processing algorithm, processing the obtained original signals to eliminate noise interference and standardize the signal format, so as to generate signal-purified data. Using a support vector machine and a random forest algorithm in the present invention enhances the fault mode identification capability and the accuracy of predicting device performance degradation trends, and substantially improves the reliability of fault prediction; the combination of a Kalman filter and a multi-level decision tree provides powerful support for the integration and analysis of multi-source data, thereby ensuring the comprehensiveness and effectiveness of decision-making support information; and using a self-organizing map network and U matrix visualization technology not only shows advantages in the aspects of data mode identification and anomaly detection, but also improves the interpretability of data analysis by means of visual image displaying.

Description

A remote monitoring method and system for aviation obstruction lights Technical Field

[0001] The present invention relates to the field of monitoring technology, and in particular to a remote monitoring method and system for aviation obstruction lights. Background Art

[0002] The monitoring sector focuses on utilizing various sensor devices, communication technologies, and data processing systems to collect, transmit, and analyze real-time status information about monitored objects. This technology plays a vital role in ensuring safe facility operation, improving efficiency, and reducing unnecessary maintenance costs. In aviation, this technology is particularly critical, as it impacts multiple aspects of flight safety and airport operations.

[0003] The remote monitoring method for aviation obstruction lights (AOL) utilizes remote technology to monitor and manage these lights. Its primary purpose is to ensure the reliable operation of these lights and promptly detect and address any faults, thereby ensuring safe flight at night or in low visibility conditions. By enabling real-time monitoring, remote control, and management of AOL lights, this approach aims to improve monitoring efficiency, reduce the need for manual inspections, and rapidly respond to any operational anomalies, ensuring that AOL lights continue to play their vital role in aviation safety.

[0004] Traditional methods for fault prediction and performance evaluation fail to utilize advanced machine learning algorithms, resulting in low fault identification and prediction accuracy. Inadequate integration and analysis of multi-source data leads to incomplete decision-making support information, hindering maintenance strategy formulation and resource optimization. The lack of effective data visualization tools also complicates the understanding and analysis of monitoring data. These deficiencies lead to inefficient monitoring, delayed maintenance responses, and inappropriate resource allocation, impacting the performance and reliability of the entire aviation obstruction light monitoring system. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a remote monitoring method and system for aviation obstruction lights.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a method for remote monitoring of aviation obstruction lights, comprising the following steps:

[0007] S1: Based on external sensors, the electromagnetic signals emitted by aviation obstruction lights are collected. Signal processing algorithms are used to process the collected raw signals, eliminate noise interference, standardize the signal format, and generate signal purification data.

[0008] S2: Based on the signal purification data, using wavelet transform and Fourier transform, perform time and frequency analysis on the signal, extract key waveform features, decompose the signal through multiple transformations, and identify and extract local time and frequency features to generate a feature waveform set;

[0009] S3: Based on the characteristic waveform set, a support vector machine is used to analyze the characteristics, compare the real-time waveform with the historical waveform, identify signals that deviate from the normal pattern, predict potential faults, and generate a fault prediction result;

[0010] S4: Based on the fault prediction results and in combination with the historical performance data of the obstruction light, a performance model is constructed using a random forest algorithm to analyze the performance degradation trend and failure probability, and the remaining life cycle of the obstruction light is predicted to generate a life cycle assessment;

[0011] S5: Based on the life cycle assessment, a Kalman filter is applied to fuse data from multiple sources to establish a data view, and a multi-level decision tree is used to analyze the fused data, analyze the data structure and key features, and generate decision support information;

[0012] S6: Based on the decision support information, a self-organizing map network is used to perform topological mapping and clustering of the data, and a U-matrix visualization technique is used to map the data features into a two-dimensional space, identify data patterns and anomalies, and generate an anomaly pattern diagram;

[0013] S7: Based on the abnormal pattern diagram and in combination with the decision support information, a hierarchical analysis process is used to analyze monitoring performance and reliability, and based on the analysis results, maintenance, fault response and resource allocation solutions are provided to generate resource optimization and maintenance plans.

[0014] As a further solution of the present invention, the signal purification data includes a signal sequence processed by Kalman filtering, a signal strength standard index and a signal frequency calibration result; the characteristic waveform set includes local time-frequency characteristics, frequency domain characteristics, and peak and trough information of the signal amplitude; the fault prediction results include abnormal waveform identification records, classification information of predicted fault types, and estimated values ​​of the probability of fault occurrence; the life cycle assessment includes a degradation curve of the obstacle light performance, a predicted maintenance time point, and an estimated value of the remaining operating time; the decision support information includes a data view, key indicator analysis results, and potential risk point prompts; the abnormal pattern diagram includes a data cluster distribution diagram, an abnormal pattern marking area, and a correlation measurement of the clustering results; the resource optimization and maintenance plan includes a scheduled maintenance schedule, a resource reconfiguration plan, and a fault response priority ranking.

[0015] As a further solution of the present invention, the electromagnetic signals emitted by aviation obstruction lights are collected based on external sensors, and the collected raw signals are processed using signal processing algorithms to eliminate noise interference and standardize the signal format. The specific steps for generating signal purification data are as follows:

[0016] S101: Based on external sensors, the original electromagnetic signal is collected, the signal is denoised using a wavelet noise reduction algorithm, and the denoised signal is subjected to time series analysis to remove environmental noise and electromagnetic interference to generate a denoised signal;

[0017] S102: Based on the denoised signal, applying Z-score normalization to normalize the signal amplitude and frequency, and performing amplitude-frequency analysis on the normalized signal to determine the consistency and comparability of the signal, thereby generating a standardized signal;

[0018] S103: Based on the standardized signal, a linear detrending algorithm is used to perform trend elimination processing on the signal, and a periodic analysis is performed on the detrended signal to eliminate non-stationary trends and optimize signal stability, thereby generating a detrended signal;

[0019] S104: Based on the detrended signal, a time series synchronization algorithm is used to perform timestamp calibration on the signal, a data format conversion tool is used to convert the data encoding of the signal, and the signal format is matched to the analysis and processing requirements to generate signal purification data.

[0020] As a further solution of the present invention, based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal to extract key waveform features, decompose the signal through multiple transformations, and identify and extract local time and frequency features. The steps of generating a characteristic waveform set are specifically as follows:

[0021] S201: Based on the signal purification data, using discrete wavelet transform, perform multi-scale time-frequency analysis on the signal, and extract local features from the analysis results to identify local time-frequency features of the signal and generate wavelet feature data;

[0022] S202: Based on the wavelet feature data, use fast Fourier transform to perform frequency domain analysis on the signal, and perform global feature extraction on the analysis result to identify the global frequency component of the signal and generate frequency domain feature data;

[0023] S203: Based on the wavelet feature data and the frequency domain feature data, applying feature cascade fusion to fuse the local time-frequency features and the global frequency features, and performing information optimization on the fused features to retain key feature information and generate comprehensive feature data;

[0024] S204: Based on the comprehensive feature data, principal component analysis is used to perform feature dimensionality reduction and optimization processing, and the information content of the reduced and optimized features is evaluated to screen the optimal features and generate a feature waveform set.

[0025] As a further solution of the present invention, based on the characteristic waveform set, a support vector machine is used to analyze the features, compare the real-time waveform and the historical waveform, identify signals that deviate from the normal pattern, and predict potential faults. The steps of generating the fault prediction result are specifically as follows:

[0026] S301: Based on the characteristic waveform set, a support vector machine is used to analyze the characteristic data and perform pattern recognition. By comparing the real-time waveform with the historical waveform data, signal characteristics that deviate from the normal waveform pattern are identified, and an abnormal signal recognition result is generated;

[0027] S302: Based on the abnormal signal recognition result, applying the K-means clustering algorithm to perform pattern grouping processing on the abnormal signal, identifying and distinguishing different types of signal abnormal patterns, and generating a fault pattern classification result;

[0028] S303: Based on the fault mode classification result, using the Apriori algorithm, performing association analysis on the classified fault modes, identifying potential associations between differentiated fault modes and frequently occurring pattern combinations, and generating a fault mode analysis result;

[0029] S304: Based on the failure mode analysis result, a Bayesian estimation method is used to perform probability analysis on the identified failure modes, perform failure probability estimation, calculate and predict the occurrence probabilities of multiple failure modes, and generate a failure prediction result.

[0030] As a further solution of the present invention, based on the fault prediction results and combined with the historical performance data of the obstruction lights, a performance model is constructed using a random forest algorithm to analyze the performance degradation trend and failure probability, and to predict the remaining life cycle of the obstruction lights. The steps for generating a life cycle assessment are specifically as follows:

[0031] S401: Based on the fault prediction result and in combination with the historical performance data of the obstacle light, a random forest algorithm is used to analyze the performance trend of the data set, identify the performance degradation trend and key points, and generate a performance degradation analysis result;

[0032] S402: Based on the performance degradation analysis results, an autoregressive integrated moving average model is used to perform a time series analysis on the performance data, and a prediction model for the performance change of the obstacle light is established to generate a performance change model;

[0033] S403: Based on the performance change model, the Cox proportional risk model is used to analyze the life of the obstruction light, perform risk and life prediction, estimate the remaining life cycle and failure probability of the obstruction light, and generate a life prediction analysis result;

[0034] S404: Based on the life prediction analysis results, the performance and remaining life of the obstruction light are analyzed using multi-criteria decision analysis, and the performance and life are evaluated to determine the status and maintenance requirements of the obstruction light and generate a life cycle assessment.

[0035] As a further solution of the present invention, based on the life cycle assessment, a Kalman filter is applied to fuse data from multiple sources to establish a data view, and a multi-level decision tree is used to analyze the fused data and analyze the data structure and key features to generate decision support information. Specifically, the steps are as follows:

[0036] S501: Based on the life cycle assessment, a Kalman filter is used to perform real-time updating and noise elimination on multi-source data, and a decision-layer fusion method is used to process data provided by multiple sensors, optimize data accuracy and uncertainty, and generate an optimized data view.

[0037] S502: Based on the optimized data view, apply random forest and decision tree algorithms to analyze the data set, vote on the results by constructing multiple decision trees, perform feature criticality assessment and data classification, and generate decision tree analysis results;

[0038] S503: Based on the decision tree analysis results, principal component analysis is used to perform dimensionality reduction processing, optimize the number of features by converting the data into a new feature space, adjust the feature correlation and compress the data to generate a structural analysis result;

[0039] S504: Based on the structural analysis results, an information integration framework and data warehouse technology are used to combine key information and features from multiple analysis stages to extract data and refine information, thereby generating decision support information.

[0040] As a further solution of the present invention, based on the decision support information, a self-organizing map network is used to perform topological mapping and clustering of the data. By using U-matrix visualization technology, the data features are mapped to a two-dimensional space, and data patterns and abnormal situations are identified. The steps of generating an abnormal pattern map are specifically as follows:

[0041] S601: Based on the decision support information, a self-organizing map network is used to perform topological mapping and feature clustering processing on the data, and pattern recognition and classification are performed through network self-learning adjustment to generate a topological mapping clustering result;

[0042] S602: Based on the topological mapping clustering result, using U matrix visualization technology, perform two-dimensional spatial mapping processing on the data features, represent the data point positions by color grayscale, and visualize the data pattern to generate a data feature visualization graph;

[0043] S603: Based on the data feature visualization graph, hierarchical density clustering is used to mine the data, and by detecting dense areas of the data, abnormal pattern recognition and normal pattern classification are performed to generate a pattern recognition graph;

[0044] S604: Based on the pattern recognition graph, a clustering quality assessment method is used to perform a quality inspection on the identified pattern, and clustering indices are calculated to perform pattern assessment and quality judgment to generate an abnormal pattern graph.

[0045] As a further solution of the present invention, based on the abnormal pattern diagram and in combination with the decision support information, a hierarchical analysis process is used to analyze monitoring performance and reliability, and based on the analysis results, maintenance, fault response and resource allocation plans are provided. The steps of generating a resource optimization and maintenance plan are specifically as follows:

[0046] S701: Based on the abnormal pattern diagram and in combination with the decision support information, a hierarchical analysis process is used to analyze the potential impact of the failure mode on performance, evaluate the monitoring performance and reliability, perform failure cause analysis and impact assessment, and generate a performance reliability assessment result;

[0047] S702: Based on the performance reliability assessment results, a risk priority number scoring method is applied to calculate a risk priority number, which includes the product of severity, probability of occurrence, and difficulty of detection, to perform quantitative risk assessment and ranking, and generate a risk priority assessment result.

[0048] S703: Based on the risk priority assessment result, a linear programming model is used to establish a linear programming problem to balance resource allocation and cost, optimize and analyze the monitoring resource configuration, and generate a resource optimization configuration result;

[0049] S704: Based on the resource optimization configuration result, particle swarm optimization is used to combine resource configuration and risk assessment results to establish a maintenance plan and fault response strategy, and resource allocation is performed to generate a resource optimization and maintenance plan.

[0050] An aviation obstruction light remote monitoring system, the aviation obstruction light remote monitoring system is used to execute the above-mentioned aviation obstruction light remote monitoring method, the system includes a signal processing module, a feature extraction module, a fault prediction module, a life cycle assessment module, a decision support module, and a resource optimization module;

[0051] The signal processing module collects raw electromagnetic signals based on external sensors, uses a wavelet noise reduction algorithm to denoise the signals, applies Z-score standardization to normalize the signal amplitude and frequency, uses a linear detrending algorithm to eliminate the trend of the signals, uses a time series synchronization algorithm to perform time stamp calibration on the signals, performs signal data encoding conversion, and generates a purified signal data set;

[0052] The feature extraction module uses discrete wavelet transform to perform multi-scale time-frequency analysis on the signal based on the purified signal data set, uses fast Fourier transform to perform frequency domain analysis on the signal and extract global features, applies feature cascade fusion to fuse local time-frequency features and global frequency features, and then performs feature dimensionality reduction and optimization through principal component analysis to generate a feature waveform set;

[0053] The fault prediction module uses a support vector machine to perform pattern recognition on the feature data based on the feature waveform set, identifies signal features that deviate from the normal waveform pattern, applies the K-means clustering algorithm to perform pattern grouping processing on abnormal signals, uses the Apriori algorithm to perform association analysis on the classified fault patterns, and uses the Bayesian estimation method to perform probability analysis on the identified fault patterns to generate fault prediction analysis results;

[0054] The life cycle assessment module is based on the results of the fault prediction analysis and combines the historical performance data of the obstruction lights. It uses the random forest algorithm to analyze the performance trend of the data set, adopts the autoregressive integral moving average model to establish a prediction model for the performance change of the obstruction lights, and uses the Cox proportional hazard model to predict the risk and life of the obstruction lights. Then, through multi-criteria decision analysis, the performance and remaining life of the obstruction lights are evaluated to generate the life cycle assessment results.

[0055] The decision support module uses a Kalman filter to perform real-time updates and noise elimination on multi-source data based on the results of life cycle assessments, applies random forest and decision tree algorithms to analyze data sets and assess feature criticality, performs data dimensionality reduction through principal component analysis, and then uses an information integration framework and data warehouse technology to extract and refine key information and features from multiple analysis stages to generate a decision support information set.

[0056] The resource optimization module is based on the decision support information set and uses a self-organizing map network to perform topological mapping and feature clustering on the data. Through U-matrix visualization technology, it performs two-dimensional spatial mapping on the data features. Hierarchical density clustering is used to mine and recognize patterns in the data. Combined with the performance evaluation results, the particle swarm optimization algorithm is used to establish maintenance plans and fault response strategies to generate resource optimization and maintenance strategies.

[0057] Compared with the prior art, the advantages and positive effects of the present invention are:

[0058] In the present invention, through the use of support vector machines and random forest algorithms, the ability to identify fault patterns and the accuracy of predicting equipment performance degradation trends are enhanced, significantly improving the reliability of fault prediction. The combination of Kalman filters and multi-level decision trees provides strong support for the integration and analysis of multi-source data, thereby ensuring the comprehensiveness and effectiveness of decision support information. The application of self-organizing mapping networks and U-matrix visualization technology not only demonstrates advantages in data pattern recognition and anomaly detection, but also improves the interpretability of data analysis through intuitive image display. In summary, this method demonstrates significant advantages in improving monitoring efficiency, accurately predicting equipment maintenance needs, and optimizing resource allocation, greatly improving the overall performance of aviation obstruction light monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] FIG1 is a schematic diagram of the workflow of the present invention;

[0060] FIG2 is a flow chart of the refinement of S1 of the present invention;

[0061] FIG3 is a flow chart of the refinement of S2 of the present invention;

[0062] FIG4 is a flow chart of the refinement of S3 of the present invention;

[0063] FIG5 is a flow chart of the refinement of S4 of the present invention;

[0064] FIG6 is a flow chart of the refinement of S5 of the present invention;

[0065] FIG7 is a flow chart of the refinement of S6 of the present invention;

[0066] FIG8 is a flow chart of the refinement of S7 of the present invention;

[0067] FIG9 is a system flow chart of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0069] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0070] Example 1

[0071] Referring to FIG1 , the present invention provides a technical solution: a method for remotely monitoring aviation obstruction lights, comprising the following steps:

[0072] S1: Based on external sensors, the electromagnetic signals emitted by aviation obstruction lights are collected. Signal processing algorithms are used to process the collected raw signals, eliminate noise interference, standardize the signal format, and generate signal purification data.

[0073] S2: Based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal, extract key waveform features, decompose the signal through multiple transformations, and identify and extract local time and frequency features to generate a feature waveform set;

[0074] S3: Based on the feature waveform set, a support vector machine is used to analyze the features, compare the real-time waveform with the historical waveform, identify signals that deviate from the normal pattern, predict potential faults, and generate fault prediction results;

[0075] S4: Based on the fault prediction results and combined with the historical performance data of the obstruction lights, a performance model is constructed using the random forest algorithm to analyze the performance degradation trend and failure probability, and to predict the remaining life cycle of the obstruction lights to generate a life cycle assessment.

[0076] S5: Based on life cycle assessment, the Kalman filter is applied to fuse data from multiple sources to establish a data view. The fused data is analyzed using a multi-level decision tree, and the data structure and key features are analyzed to generate decision support information.

[0077] S6: Based on decision support information, a self-organizing map network is used to perform topological mapping and clustering of data. Using U-matrix visualization technology, data features are mapped into a two-dimensional space, and data patterns and anomalies are identified to generate an anomaly pattern diagram.

[0078] S7: Based on the abnormal pattern diagram and combined with decision support information, a hierarchical analysis process is used to analyze monitoring performance and reliability. Based on the analysis results, maintenance, fault response and resource allocation plans are provided to generate resource optimization and maintenance plans.

[0079] The signal purification data includes the signal sequence processed by Kalman filtering, the standard index of signal strength and the signal frequency calibration results. The characteristic waveform set includes the local time-frequency characteristics, frequency domain characteristics, and the peak and trough information of the signal amplitude. The fault prediction results include the abnormal waveform identification records, the classification information of the predicted fault type, and the estimated value of the fault occurrence probability. The life cycle assessment includes the degradation curve of the obstacle light performance, the predicted maintenance time point, and the estimated value of the remaining operating time. The decision support information includes data views, key indicator analysis results, and potential risk point prompts. The abnormal pattern diagram includes the data cluster distribution diagram, the abnormal pattern marking area, and the correlation measurement of the clustering results. The resource optimization and maintenance plan includes the scheduled maintenance schedule, the resource reconfiguration plan, and the fault response priority ranking.

[0080] In step S1, external sensors are used to collect electromagnetic signals emitted by aviation obstruction lights. The collected original signals are stored in a time series data format and contain noise and interference. To ensure the quality and reliability of the signal, a signal processing algorithm is used to pre-process the original signal, including filtering technology to eliminate noise and signal normalization operation to unify the signal format. The filtering technology selects appropriate filters based on the characteristics of the signal, such as low-pass, high-pass or band-pass filters, to remove non-related frequency components. The signal normalization operation ensures that the signal is processed and analyzed under the same standard, which facilitates the execution of subsequent steps. After this series of processing, the generated signal purification data includes the signal sequence processed by Kalman filtering, the signal strength standard index and the signal frequency calibration result, which provides high-quality input data for subsequent feature extraction and analysis.

[0081] In step S2, based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal. Wavelet transform and Fourier transform are two powerful signal processing tools that can reveal the characteristics of the signal from different angles. Wavelet transform extracts the local time-frequency characteristics of the signal through multi-scale analysis technology, which is particularly suitable for feature extraction of non-stationary signals. Fourier transform converts the signal to the frequency domain and extracts its frequency characteristics. By combining these two transforms, more comprehensive features can be extracted from the signal, including local time-frequency characteristics, frequency domain characteristics, and peak and trough information of the signal amplitude. These features constitute a characteristic waveform set, which provides an important basis for further analysis of the signal and fault prediction.

[0082] In step S3, based on the feature waveform set, a support vector machine is used to analyze the features and predict faults. The support vector machine is a powerful supervised learning algorithm that can handle complex classification tasks. In this step, the support vector machine is used to compare and analyze real-time waveforms and historical waveforms to identify signals that deviate from normal patterns. By learning the characteristics of normal and abnormal signals in historical data, the support vector machine can predict the occurrence of potential faults. The generated fault prediction results include abnormal waveform identification records, classification information of predicted fault types, and estimated values ​​of the probability of fault occurrence. This step is crucial for timely identification and prevention of aviation obstruction light failures.

[0083] In step S4, a performance model is constructed using the random forest algorithm based on the fault prediction results and the historical performance data of the obstruction lights. The random forest algorithm is an integrated learning algorithm that improves the accuracy and robustness of the model by constructing multiple decision trees and combining their prediction results. In this step, the random forest algorithm analyzes the performance degradation trend and failure probability of the obstruction lights and predicts the remaining life cycle of the obstruction lights. The generated life cycle assessment includes the degradation curve of the obstruction light performance, the predicted maintenance time point and the estimated remaining operating time. This step plays an important role in formulating maintenance plans and ensuring the reliable operation of obstruction lights.

[0084] In step S5, based on the life cycle assessment, the Kalman filter is applied to fuse and analyze the data. The Kalman filter is an effective data fusion tool that can combine data from different sources to provide the best estimate of the system status. In this step, the Kalman filter fuses the life cycle assessment and other related data to establish a unified data view. Subsequently, a multi-level decision tree is used to analyze the fused data, analyze the data structure and key features, and generate decision support information including data views, key indicator analysis results and potential risk point prompts. This step is crucial for in-depth understanding of the operating status of the obstacle lights and making reasonable decisions.

[0085] In step S6, based on decision support information, a self-organizing map network is used to perform topological mapping and clustering on the data. The self-organizing map network is an unsupervised learning algorithm that can map high-dimensional data to a low-dimensional space and retain the topological structure of the data in the process. Through U-matrix visualization technology, data features are mapped to a two-dimensional space to form an intuitive image representation. This image representation reveals the intrinsic structure of the data, including data cluster distribution maps, abnormal pattern marking areas, and correlation measurements of clustering results. This information helps to identify data patterns and abnormal situations, providing important support for fault diagnosis and system maintenance.

[0086] In step S7, based on the abnormal pattern diagram and decision support information, the hierarchical analysis process is used to conduct a comprehensive analysis of the monitoring performance and reliability. The hierarchical analysis process is a multi-criteria decision analysis tool that solves complex decision-making problems by constructing a hierarchical model and comparing and weighting the elements in the model. In this step, the hierarchical analysis process is used to analyze the performance and reliability of the obstacle light monitoring system, as well as its impact on maintenance, fault response and resource allocation plans. Based on the analysis results, a resource optimization and maintenance plan is generated. The plan includes a scheduled maintenance schedule, a resource reconfiguration plan and a fault response priority ranking. This step is of great significance for improving the efficiency and reliability of the monitoring system and ensuring the continuous and stable operation of the system.

[0087] Refer to Figure 2. Using external sensors to collect electromagnetic signals from aviation obstruction lights, a signal processing algorithm is used to process the collected raw signals, eliminate noise interference, and standardize the signal format. The specific steps for generating signal purification data are as follows:

[0088] S101: Based on external sensors, the original electromagnetic signal is collected, the signal is denoised using a wavelet noise reduction algorithm, and the denoised signal is subjected to time series analysis to remove environmental noise and electromagnetic interference to generate a denoised signal;

[0089] S102: Based on the denoised signal, Z-score normalization is applied to normalize the signal amplitude and frequency, and amplitude-frequency analysis is performed on the normalized signal to determine the consistency and comparability of the signal, thereby generating a standardized signal;

[0090] S103: Based on the standardized signal, a linear detrending algorithm is used to perform trend elimination processing on the signal, and a periodic analysis is performed on the detrended signal to eliminate non-stationary trends and optimize signal stability, thereby generating a detrended signal;

[0091] S104: Based on the detrended signal, a time series synchronization algorithm is used to perform timestamp calibration on the signal. A data format conversion tool is used to convert the signal data encoding and match the signal format to the analysis and processing requirements to generate signal purification data.

[0092] In sub-step S101, the original electromagnetic signal collected by the external sensor is first stored in analog or digital form. This data format depends on the sensor type and the characteristics of the signal. The signal is denoised using a wavelet denoising algorithm. This process involves selecting an appropriate wavelet basis and decomposition level to decompose the signal. Then, the noise threshold of each decomposition level is estimated and the wavelet coefficients are soft-thresholded according to these thresholds to eliminate the noise component. The denoised signal is then analyzed through time series analysis, mainly using autocorrelation and partial autocorrelation functions to identify and eliminate the effects of environmental noise and electromagnetic interference, thereby generating a denoised signal. The signal is stored in a time series data format and is represented as a mapping of time points and signal strengths for subsequent processing.

[0093] In sub-step S102, based on the denoised signal, the Z-score normalization algorithm is applied to convert the data into a form with zero mean and unit variance by calculating the mean and standard deviation of each data point, thereby achieving normalization of the signal amplitude and frequency. The normalized signal is subjected to amplitude-frequency analysis, usually using fast Fourier transform, to extract the frequency components of the signal and analyze their consistency and comparability. The standardized signal generated by this process is stored in the frequency domain data format, which reflects the intensity and distribution of the signal in each frequency band, which is helpful for subsequent feature recognition and classification.

[0094] In sub-step S103, based on the standardized signal, a linear detrending algorithm is used to identify and eliminate the linear trend components in the signal through a linear regression model, so that the signal tends to be stable. The detrended signal is subjected to periodic analysis, and a periodogram or autocorrelation function is usually applied to identify the periodic components of the signal. The purpose of this step is to eliminate non-stationary trends and optimize the stability of the signal. The generated detrended signal is stored in an adjusted time series format to provide more accurate signal characteristics for subsequent analysis.

[0095] In sub-step S104, based on the detrended signal, a time series synchronization algorithm is used to perform timestamp calibration on the signal to ensure the synchronization consistency of the signal data with the actual event. The data format conversion tool is used to convert the signal data into a format suitable for analysis and processing, such as from analog signal to digital signal or vice versa. In addition, signal format matching ensures that the signal data meets the analysis and processing requirements. Through this series of processing, the generated signal purification data is presented in an optimized time series format, providing a clear and accurate basis for the next step of feature extraction and data analysis.

[0096] Please refer to Figure 3. Based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal to extract key waveform features. The signal is decomposed through multiple transformations, and local time and frequency features are identified and extracted. The specific steps for generating a feature waveform set are as follows:

[0097] S201: Based on the signal purification data, a discrete wavelet transform is used to perform multi-scale time-frequency analysis on the signal, and local features are extracted from the analysis results to identify the local time-frequency features of the signal and generate wavelet feature data;

[0098] S202: Based on the wavelet feature data, use fast Fourier transform to perform frequency domain analysis on the signal, and perform global feature extraction on the analysis result to identify the global frequency component of the signal and generate frequency domain feature data;

[0099] S203: Based on the wavelet feature data and the frequency domain feature data, feature cascade fusion is applied to fuse the local time-frequency features and the global frequency features, and information optimization is performed on the fused features to retain key feature information and generate comprehensive feature data;

[0100] S204: Based on the comprehensive feature data, principal component analysis is used to perform feature dimensionality reduction and optimization processing, and the information content of the reduced and optimized features is evaluated to screen the optimal features and generate a feature waveform set.

[0101] In sub-step S201, based on the signal purification data, discrete wavelet transform (DWT) is used to perform multi-scale time-frequency analysis. The signal purification data exists in the form of a time series, and each data point represents the signal intensity at a time point. Discrete wavelet transform is an analysis method that can provide time and frequency information at the same time. It is very suitable for feature extraction of non-stationary signals. In this process, the signal is first decomposed into sub-band signals of different scales, and each sub-band corresponds to a specific frequency range of the signal. Then, local features are extracted for each sub-band signal, including the energy, mean and standard deviation of the sub-band. These local features can reveal the behavior of the signal within a specific time and frequency range. Wavelet feature data is generated through multi-scale analysis and local feature extraction of discrete wavelet transform. The data reflects the local time-frequency characteristics of the signal, providing a basis for subsequent frequency domain analysis and feature fusion.

[0102] In sub-step S202, based on the wavelet feature data, the signal is subjected to frequency domain analysis using the fast Fourier transform (FFT). The fast Fourier transform is a computationally efficient Fourier transform method used to convert time domain signals into frequency domain signals. In this process, the FFT analyzes the overall frequency components of the signal and reveals the frequency characteristics of the signal over the entire time range. By performing global feature extraction on the analysis results, including the main frequency components, frequency distribution, and spectral density of the signal, frequency domain feature data is generated. The frequency domain feature data provides information about the global frequency components of the signal, providing a basis for subsequent feature fusion and optimization.

[0103] In sub-step S203, based on the wavelet feature data and the frequency domain feature data, feature cascade fusion is applied to fuse the local time-frequency features and the global frequency features. Feature cascade fusion is a feature integration technology that combines features from different sources to form a comprehensive feature representation. In this process, the local time-frequency features and the global frequency features are combined to generate a more comprehensive signal feature representation. Then, the fused features are subjected to information optimization processing, including feature selection and feature weighting, to retain the most critical feature information and eliminate redundant or unimportant features. Through feature cascade fusion and information optimization processing, comprehensive feature data is generated. The data integrates the local time-frequency features and the global frequency features, providing rich information for further analysis and identification of the signal.

[0104] In sub-step S204, based on the comprehensive feature data, principal component analysis (PCA) is used to perform feature dimensionality reduction and optimization processing. Principal component analysis is a statistical method used to reduce the dimension of data while retaining the variability of the original data as much as possible. In this process, PCA represents the original features by constructing new orthogonal features (principal components). The new features are sorted according to the degree to which they can explain the variability of the original data. Then, the information content of the reduced and optimized features is evaluated, and the principal components that can most effectively represent the signal characteristics are selected. In this way, PCA can reduce the number of features and simplify subsequent analysis and processing processes while retaining the most critical feature information. The generated feature waveform set includes the optimal features that have undergone dimensionality reduction and optimization processing, providing high-quality feature representation for the final identification and analysis of the signal.

[0105] Refer to Figure 4. Based on the feature waveform set, a support vector machine is used to analyze the features, compare the real-time waveform with the historical waveform, identify signals that deviate from the normal pattern, and predict potential faults. The specific steps for generating the fault prediction result are as follows:

[0106] S301: Based on the feature waveform set, a support vector machine is used to analyze the feature data and perform pattern recognition. By comparing the real-time waveform with the historical waveform data, signal features that deviate from the normal waveform pattern are identified, and an abnormal signal recognition result is generated;

[0107] S302: Based on the abnormal signal recognition results, apply the K-means clustering algorithm to perform pattern grouping processing on the abnormal signals, identify and distinguish different types of signal abnormal patterns, and generate a fault mode classification result;

[0108] S303: Based on the fault mode classification results, the Apriori algorithm is used to perform correlation analysis on the classified fault modes, identify potential correlations between differentiated fault modes and frequently occurring pattern combinations, and generate fault mode analysis results;

[0109] S304: Based on the failure mode analysis results, a Bayesian estimation method is used to perform probability analysis on the identified failure modes, perform failure probability estimation, calculate and predict the occurrence probabilities of multiple failure modes, and generate failure prediction results.

[0110] In sub-step S301, based on the feature waveform set, the support vector machine (SVM) algorithm is used for pattern recognition. The feature waveform set exists in the form of multi-dimensional data, and each dimension represents a feature of the waveform. The support vector machine is a supervised learning algorithm used to identify patterns and trends in data. In this step, the SVM algorithm classifies the feature data in a high-dimensional space by constructing one or more hyperplanes. The algorithm first learns the normal pattern in the historical waveform data, and then compares the real-time waveform with the learned pattern. Through this comparison, the SVM can identify signal features that deviate from the normal waveform pattern. This deviation may be caused by a failure or performance degradation of the aviation obstruction light. The generated abnormal signal recognition results include the normal or abnormal state of each waveform, as well as a quantitative indicator of the degree of abnormality. The results provide a basis for subsequent fault mode grouping processing.

[0111] In sub-step S302, based on the abnormal signal recognition results, the K-means clustering algorithm is applied to perform pattern grouping processing on the abnormal signals. K-means is a simple and effective clustering algorithm that classifies data by minimizing the distance from each point to the center of the cluster to which it belongs. In this step, the algorithm first determines the number of clusters, and then inputs the abnormal signal feature data into the K-means algorithm. The algorithm adjusts the position of the cluster center through an iterative process until the best classification effect is achieved. In this way, the algorithm can identify and distinguish different types of signal abnormal patterns. The generated fault mode classification results include the cluster label of each abnormal signal and the feature center of each cluster. The results provide a basis for identifying potential associations between differentiated fault modes.

[0112] In sub-step S303, based on the fault mode classification results, the Apriori algorithm is used to perform association analysis. The Apriori algorithm is a commonly used association rule learning algorithm used to mine interesting relationships between items from large amounts of data. In this step, the fault mode classification results are used as input data. The Apriori algorithm mines association rules between fault modes by identifying frequently occurring item sets. The algorithm first generates a set of candidate items, then calculates the frequency of these sets in the data, and filters out frequent item sets using a predefined support threshold. In this way, the algorithm can identify potential associations between differentiated fault modes and frequently occurring pattern combinations. The generated fault mode analysis results include association rules and support information between fault modes. The results provide important reference information for further failure probability estimation.

[0113] In sub-step S304, based on the results of the failure mode analysis, the Bayesian estimation method is used to perform a probability analysis on the identified failure modes. Bayesian estimation is a statistical method that updates the probability of an event by considering prior knowledge and new evidence. In this step, the failure mode analysis results are used as input data. The Bayesian estimation method calculates and predicts the probability of occurrence of multiple failure modes by combining prior knowledge (such as historical failure data) and new failure mode analysis results. This process involves calculating conditional probabilities and marginal probabilities, and ultimately generates the probability of occurrence of each failure mode. The generated fault prediction results include probability estimates for each failure mode, as well as a comprehensive evaluation of the occurrence of the failure. These results play an important role in predicting and preventing possible future failures, and more effectively formulating maintenance strategies and optimizing resource allocation.

[0114] Refer to Figure 5. Based on the failure prediction results and the historical performance data of the obstruction lights, a performance model is constructed using the random forest algorithm to analyze the performance degradation trend and failure probability, and to predict the remaining life cycle of the obstruction lights. The specific steps for generating a life cycle assessment are as follows:

[0115] S401: Based on the fault prediction results and combined with the historical performance data of the obstruction lights, the random forest algorithm is used to analyze the performance trend of the data set, identify the performance degradation trend and key points, and generate the performance degradation analysis results;

[0116] S402: Based on the performance degradation analysis results, an autoregressive integrated moving average model is used to perform time series analysis on the performance data, and a prediction model for the performance change of the obstacle light is established to generate a performance change model;

[0117] S403: Based on the performance change model, the Cox proportional risk model is used to analyze the life of the obstruction lights, perform risk and life prediction, estimate the remaining life cycle and failure probability of the obstruction lights, and generate life prediction analysis results;

[0118] S404: Based on the life prediction analysis results, use multi-criteria decision analysis to analyze the performance and remaining life of the obstacle light, evaluate the performance and life, determine the status and maintenance requirements of the obstacle light, and generate a life cycle assessment.

[0119] In sub-step S401, based on the fault prediction results and combined with the historical performance data of the obstacle lights, the random forest algorithm is used to conduct an in-depth analysis of the performance trends of the data set. The random forest algorithm is an integrated learning method that improves the accuracy and robustness of the overall model by constructing multiple decision trees and combining their prediction results. In this process, each decision tree is trained on a randomly selected data subset and feature subset. This method can effectively reduce the risk of overfitting of the model. By analyzing historical performance data and real-time fault prediction data, the random forest model can identify and quantify the degradation trend and key change points of the obstacle light performance. The generated performance degradation analysis results describe in detail the performance trend of the obstacle light over time, highlight the key stages of performance degradation and potential risk points, and provide a decision-making basis for the maintenance and repair of the obstacle light.

[0120] In sub-step S402, based on the performance degradation analysis results, the autoregressive integrated moving average (ARIMA) model is used to perform time series analysis on the performance data. The ARIMA model is a statistical model widely used in time series prediction. It combines autoregressive (AR), differencing (I) and moving average (MA) components to describe the characteristics of time series data. In this step, the model first determines the stability of the time series and converts the unstable series into a stable series through differential transformation. Then, the model fits the time series by identifying the autoregressive and moving average terms in the data, and finally establishes a model for predicting future performance changes. The generated performance change model can predict the future performance trend of the obstacle light, providing a scientific basis for predicting performance degradation in advance and taking preventive measures.

[0121] In sub-step S403, based on the performance change model, the Cox proportional hazard model is used to analyze the life of the obstruction light. The Cox proportional hazard model is a statistical method used for survival analysis, which can evaluate the impact of multiple variables on survival time. In this process, the model takes into account the performance change data of the obstruction light and other factors that may affect the life. By constructing a risk function to describe the change in the risk of obstruction light failure over time, the remaining life cycle of the obstruction light and the failure probability under different conditions can be estimated through model analysis. The generated life prediction analysis results provide scientific data support for the maintenance planning and life cycle management of the obstruction light.

[0122] In sub-step S404, based on the results of the life prediction analysis, a comprehensive evaluation of the performance and remaining life of the obstruction lights is performed using multi-criteria decision analysis. Multi-criteria decision analysis is a decision support tool that helps make optimal decisions by comprehensively considering multiple influencing factors and decision criteria. In this step, the model comprehensively analyzes the performance degradation trend of the obstruction lights, the remaining life prediction results, and other relevant operation and maintenance data. Through this comprehensive analysis, the current status of the obstruction lights can be accurately evaluated, and reasonable judgments can be made on the need for maintenance and replacement. The generated life cycle assessment results include scores for the performance and life of the obstruction lights, status classification, and maintenance or replacement recommendations, providing strong decision support for the maintenance management and resource optimization of the obstruction lights.

[0123] Refer to Figure 6. Based on life cycle assessment, the steps for applying the Kalman filter, fusing data from multiple sources, establishing a data view, and using a multi-level decision tree to analyze the fused data and analyze the data structure and key features to generate decision support information are as follows:

[0124] S501: Based on life cycle assessment, a Kalman filter is used to perform real-time updates and noise elimination on multi-source data. A decision-layer fusion method is used to process data provided by multiple sensors, optimize data accuracy and uncertainty, and generate an optimized data view.

[0125] S502: Based on the optimized data view, the random forest and decision tree algorithms are applied to analyze the data set. The results are voted on by constructing multiple decision trees, and feature criticality assessment and data classification are performed to generate decision tree analysis results.

[0126] S503: Based on the decision tree analysis results, principal component analysis is used to perform dimensionality reduction processing, optimize the number of features by converting the data into a new feature space, adjust the feature correlation and compress the data to generate a structural analysis result;

[0127] S504: Based on the results of structural analysis, the information integration framework and data warehouse technology are used to extract data and refine information by merging key information and features from multiple analysis stages to generate decision support information.

[0128] In sub-step S501, multi-source data is processed through a Kalman filter to achieve real-time updates and noise elimination. The Kalman filter is an effective recursive filter used to estimate the state of a linear dynamic system. In this process, the initial state estimate and error covariance are first set based on the prior knowledge of the system. Then, whenever new data is input, the filter continuously adjusts the state estimate by predicting the next state and updating the current state estimate, combined with the new observation data. This processing method can effectively reduce the noise introduced by factors such as sensor errors and environmental changes, thereby improving the accuracy of the data. The decision-making layer fusion method further processes the data provided by multiple sensors, optimizes data accuracy and uncertainty, and improves the reliability and representativeness of the overall data by weighting and fusing data from different sources. The generated optimized data view provides a more accurate and comprehensive data representation, which is helpful for subsequent analysis and decision-making processes.

[0129] In sub-step S502, random forest and decision tree algorithms are applied to conduct in-depth analysis of the optimized data view. Random forest is an integrated learning method that constructs multiple decision trees and lets them vote on the final result. In this process, multiple sample subsets are first randomly selected from the original data set, and decision trees are established for each subset. These decision trees consider different feature subsets during the construction process to increase the diversity of the model. Each tree votes on the classification result of the data set, and finally the classification of the data point is determined by the category with the most votes. In addition, feature criticality evaluation is also performed to identify the features that have the greatest impact on the classification results. In this way, random forest can provide highly accurate data classification. The generated decision tree analysis results are stored in a tree-structured data format, which contains key features and classification decision paths.

[0130] In sub-step S503, principal component analysis is used for dimensionality reduction based on the results of decision tree analysis. Principal component analysis is a statistical method that converts original data into a set of linearly uncorrelated feature spaces through orthogonal transformation. In this process, the focus is on the main variation direction of the data. First, the covariance matrix of the data set is calculated, and then the eigenvalues ​​and eigenvectors of this matrix are found. These eigenvectors constitute a new feature space. The projection of the data points in this space is the principal component. The eigenvectors corresponding to the largest eigenvalues ​​are selected as principal components to achieve dimensionality reduction. This process not only reduces the complexity of the data, but also helps to reveal the structural relationship within the data. The generated structural analysis results are stored in the form of principal component scores, providing a more concise and informative data representation.

[0131] In sub-step S504, the information integration framework and data warehouse technology are used to process the structural analysis results. The information integration framework provides a systematic approach to merge key information and features from multiple analysis stages. In this process, key information is first extracted from different analysis results, such as important features, decision rules and patterns. Then, data warehouse technology is used to integrate this information to form a unified, multi-dimensional data structure. This data structure is organized in the form of data cubes, allowing users to query and analyze data from different angles and levels. The decision support information finally generated provides a comprehensive and in-depth data insight, which helps guide actual operations and decisions. The information is stored in the form of optimized and summarized data, providing powerful data support and analysis tools.

[0132] Referring to Figure 7, based on decision support information, a self-organizing map network is used to perform topological mapping and clustering of data. Using U-matrix visualization technology, data features are mapped to a two-dimensional space, and data patterns and anomalies are identified. The specific steps for generating an anomaly pattern map are as follows:

[0133] S601: Based on the decision support information, a self-organizing map network is used to perform topological mapping and feature clustering processing on the data. Through network self-learning adjustment, pattern recognition and classification are performed to generate a topological mapping clustering result.

[0134] S602: Based on the topological mapping clustering results, the U matrix visualization technology is used to perform two-dimensional spatial mapping processing on the data features, the data point positions are represented by color grayscale, and the data patterns are visualized to generate a data feature visualization graph;

[0135] S603: Based on the data feature visualization graph, hierarchical density clustering is used to mine the data. By detecting dense areas of the data, abnormal pattern recognition and normal pattern classification are performed to generate a pattern recognition graph;

[0136] S604: Based on the pattern recognition graph, a clustering quality assessment method is used to perform a quality inspection on the identified pattern. By calculating the clustering index, pattern assessment and quality judgment are performed to generate an abnormal pattern graph.

[0137] In sub-step S601, the decision support information is processed through a self-organizing map network to achieve topological mapping and feature clustering of the data. The self-organizing map network is an unsupervised neural network that maps high-dimensional data to a low-dimensional space (usually two-dimensional) through a network self-learning adjustment process while maintaining the topological structure of the data. In this process, each node of the network represents a clustering center. The network selects the node closest to the input data through a competitive learning process and adjusts the weights of the node and its neighboring nodes to make them closer to the input data. Through iterative processing, the network gradually forms a topological structure diagram that reflects the characteristics of the data, achieving pattern recognition and classification. The generated topological mapping clustering result is represented by a two-dimensional graph, where each node represents a data cluster and adjacent nodes represent data clusters with similar characteristics, providing an intuitive view for further analysis and understanding of the data.

[0138] In sub-step S602, the topological mapping clustering results are mapped into two-dimensional space using U-matrix visualization technology. U-matrix is ​​a method for displaying the similarity between nodes in a self-organizing mapping network. The location of data points and the distance between nodes are represented by color grayscale. In this process, the color depth of each node represents the similarity or difference between the node and the adjacent nodes. Dark colors represent large differences, and light colors represent small differences. This visualization method makes the recognition and interpretation of data patterns intuitive, and helps to identify potential data clusters or outliers. The generated data feature visualization diagram is presented in the form of a two-dimensional chart, providing a visually direct way to understand and interpret data features and patterns.

[0139] In sub-step S603, based on the data feature visualization diagram, hierarchical density clustering is used to conduct in-depth data mining. Hierarchical density clustering is a density-based clustering method that identifies clusters by detecting dense areas of data. In this process, the algorithm first calculates the local density of data points, and then assigns data points to the nearest high-density area according to the density value to form clusters. This method is particularly suitable for identifying clusters of irregular shapes or sizes, and can effectively identify abnormal patterns and normal patterns. The generated pattern recognition diagram is a graphical representation of the clustering results, in which different clusters are displayed with different colors or marks, which intuitively shows the structure and pattern of the data, helping users understand and analyze the inherent characteristics of the data.

[0140] In sub-step S604, based on the pattern recognition graph, a clustering quality assessment method is used to perform a quality check on the identified pattern. Clustering quality assessment usually includes calculating various clustering indices, such as the silhouette coefficient, the Davies-Bouldin index, or the Calinski-Harabasz index. These indices evaluate the quality of clustering by measuring the compactness within the cluster and the separation between clusters. Through these calculations, the quality of the clustering results can be objectively evaluated, and high-quality or suboptimal clusters can be identified. The generated abnormal pattern graph is represented in a visual form of the clustering quality assessment results, highlighting clusters with higher or lower quality, providing important reference information for further data analysis and decision-making.

[0141] Refer to Figure 8. Based on the abnormal pattern diagram and decision support information, a hierarchical analysis process is used to analyze monitoring performance and reliability. Based on the analysis results, maintenance, fault response, and resource allocation plans are provided. The specific steps for generating resource optimization and maintenance plans are as follows:

[0142] S701: Based on the abnormal pattern diagram and combined with decision support information, a hierarchical analysis process is used to analyze the potential impact of failure modes on performance, evaluate monitoring performance and reliability, conduct failure cause analysis and impact assessment, and generate performance reliability assessment results.

[0143] S702: Based on the performance reliability assessment results, apply the risk priority number scoring method to calculate the risk priority number, including the product of severity, probability of occurrence, and detection difficulty, to perform quantitative risk assessment and ranking, and generate the risk priority assessment results;

[0144] S703: Based on the risk priority assessment results, a linear programming model is used to establish a linear programming problem to balance resource allocation and cost, optimize and analyze the monitoring resource configuration, and generate resource optimization configuration results;

[0145] S704: Based on the resource optimization configuration results, particle swarm optimization is used, combined with resource configuration and risk assessment results, to establish a maintenance plan and fault response strategy, and to allocate resources to generate a resource optimization and maintenance plan.

[0146] In sub-step S701, a failure mode analysis is performed through the hierarchical analysis process combined with decision support information and abnormal pattern diagrams. The hierarchical analysis process is a decision support tool that breaks down complex problems into smaller, more manageable parts by establishing a multi-level structure. First, the various aspects of the problem are identified and defined, such as failure modes, influencing factors, and performance indicators. Then, these aspects are paired and compared to evaluate their relative importance to the overall goal. In this way, the potential impact of failure modes on monitoring performance and reliability can be quantified, and the causes of failures and their impacts can be evaluated. This process uses a specific judgment matrix and weight calculation method to ensure the consistency and reliability of the evaluation results. The generated performance reliability evaluation results are presented in the form of a structured report, which provides an in-depth understanding of the potential failures and performance impacts of the monitoring system, and provides a basis for formulating maintenance and improvement measures.

[0147] In sub-step S702, the risk priority number scoring method is applied to perform a quantitative risk assessment based on the performance reliability assessment results. The risk priority number scoring method is a quantitative tool used to evaluate and compare the severity, probability of occurrence, and difficulty of detection of risks. First, the scoring criteria of these three dimensions are defined for each failure mode, and each dimension is scored. Then, the risk priority number is calculated, which is the product of the scores of these three dimensions. Through this method, the risk levels of different failure modes can be quantitatively compared and ranked. The generated risk priority assessment results are presented in the form of tables or charts, which clearly show the risk levels of different failure modes and provide an important basis for formulating risk response measures.

[0148] In sub-step S703, a linear programming model is used to optimize the monitoring resource configuration. Linear programming is a mathematical method used to optimize a linear objective function under a series of linear constraints. First, a linear programming problem including resource allocation and cost is established. In this model, the total amount of resources, the cost of different types of resources, and the goal of resource configuration are defined, such as minimizing total cost or maximizing performance benefits. Then, a linear programming algorithm, such as the simplex method or the interior point method, is used to solve this optimization problem. In this way, resource allocation and cost can be effectively balanced, and resource configuration can be optimized. The generated resource optimization configuration results are presented in the form of an optimization report, which indicates the optimal resource allocation plan and provides guidance for improving the efficiency and effectiveness of the monitoring system.

[0149] In sub-step S704, the particle swarm optimization algorithm is used to combine resource allocation and risk assessment results to establish a maintenance plan and fault response strategy. Particle swarm optimization is an optimization algorithm based on swarm intelligence. It finds the optimal solution by simulating the social behavior of bird flocks or fish schools. In this process, a group of particles is defined, each particle represents a potential solution, such as a specific resource configuration or maintenance plan. These particles move in the solution space, adjust their positions based on their own experience and the experience of the group, and gradually move closer to the optimal solution. Through this method, particle swarm optimization can efficiently search a large range of solution space and find the best resource configuration and maintenance strategy. The generated resource optimization and maintenance plan is presented in the form of a detailed plan document, which provides specific maintenance activities and resource allocation plans for the monitoring system, providing practical guidance for system operation and maintenance and risk management.

[0150] Please refer to FIG9 , which shows a remote monitoring system for aviation obstruction lights. The remote monitoring system for aviation obstruction lights is used to implement the above-mentioned remote monitoring method for aviation obstruction lights. The system includes a signal processing module, a feature extraction module, a fault prediction module, a life cycle assessment module, a decision support module, and a resource optimization module.

[0151] The signal processing module collects raw electromagnetic signals based on external sensors, denoises the signals using a wavelet noise reduction algorithm, normalizes the signal amplitude and frequency using a Z-score, eliminates the trend of the signals using a linear detrending algorithm, calibrates the signal timestamps using a time series synchronization algorithm, and performs signal data encoding conversion to generate a clean signal dataset.

[0152] The feature extraction module uses discrete wavelet transform to perform multi-scale time-frequency analysis on the signal based on the purified signal data set. It also uses fast Fourier transform to perform frequency domain analysis on the signal and extract global features. It then applies feature cascade fusion to fuse local time-frequency features with global frequency features. It then uses principal component analysis to perform feature dimensionality reduction and optimization to generate a feature waveform set.

[0153] The fault prediction module uses a support vector machine to perform pattern recognition on the feature data based on the characteristic waveform set, identifying signal features that deviate from the normal waveform pattern. It then applies the K-means clustering algorithm to group the abnormal signals into patterns, uses the Apriori algorithm to perform correlation analysis on the classified fault patterns, and uses the Bayesian estimation method to perform probability analysis on the identified fault patterns to generate fault prediction analysis results.

[0154] The life cycle assessment module uses the random forest algorithm to analyze the performance trends of the data set based on the results of the fault prediction analysis and the historical performance data of the obstruction lights. It also uses the autoregressive integral moving average model to establish a prediction model for the performance changes of the obstruction lights. The Cox proportional hazards model is used to predict the risk and life of the obstruction lights. Then, through multi-criteria decision analysis, the performance and remaining life of the obstruction lights are evaluated to generate the life cycle assessment results.

[0155] Based on the results of life cycle assessment, the decision support module uses Kalman filters to perform real-time updates and noise elimination on multi-source data. It also uses random forest and decision tree algorithms to analyze data sets and assess feature criticality. Principal component analysis is used to reduce data dimensionality. Then, using an information integration framework and data warehouse technology, it extracts and refines key information and features from multiple analysis stages to generate a decision support information set.

[0156] The resource optimization module is based on the decision support information set and uses the self-organizing map network to perform topological mapping and feature clustering on the data. Through the U-matrix visualization technology, the data features are mapped in two-dimensional space. Hierarchical density clustering is used to mine and recognize patterns on the data. Combined with the performance evaluation results, the particle swarm optimization algorithm is used to establish maintenance plans and fault response strategies to generate resource optimization and maintenance strategies.

[0157] The signal processing module uses technologies such as wavelet noise reduction, Z-score normalization, linear detrending and time series synchronization to greatly improve the quality and reliability of the signal, ensure the accuracy and stability of the data basis of the monitoring system, and provide reliable input for subsequent analysis and decision-making.

[0158] The feature extraction module extracts accurate and comprehensive features from signals through discrete wavelet transform and fast Fourier transform, combined with feature cascade fusion and principal component analysis. This efficient feature extraction and optimization mechanism enables the system to accurately identify and respond to various signal patterns, enhancing the system's ability to identify and predict faults.

[0159] The fault prediction module integrates support vector machines, K-means clustering, Apriori algorithm and Bayesian estimation method, providing a multi-dimensional fault prediction framework. This not only improves the accuracy of fault identification, but also makes fault prediction more forward-looking and reliable. Through detailed fault analysis, the system can promptly detect and warn of potential obstruction light problems, so that maintenance measures can be taken in advance to reduce the occurrence of faults.

[0160] The Life Cycle Assessment module provides powerful data support for performance evaluation and life prediction of obstruction lights by using algorithms such as random forest, ARIMA model, and Cox proportional hazards model. This enables the maintenance team to develop more scientific and reasonable maintenance plans based on detailed data analysis, optimize the allocation of maintenance resources, extend the service life of obstruction lights, and thus reduce long-term operation and maintenance costs.

[0161] The decision support module and resource optimization module further enhance the system's decision-making ability and resource allocation efficiency. The decision support module provides precise data analysis and decision support through technologies such as Kalman filter, random forest and principal component analysis, helping the management team to make more scientific and effective decisions. The resource optimization module uses technologies such as self-organizing map network and hierarchical density clustering to achieve effective data mining and resource optimization allocation. Through the application of these advanced algorithms, the system not only improves operation and maintenance efficiency, but also significantly enhances the overall performance and reliability of the aviation obstruction lighting system.

[0162] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A remote monitoring method for aviation obstruction lights, characterized in that: The following steps are involved: Based on external sensors, the electromagnetic signals emitted by aviation obstruction lights are collected. Signal processing algorithms are used to process the collected raw signals, eliminate noise interference, standardize the signal format, and generate signal purification data. Based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal to extract key waveform features, decompose the signal through multiple transformations, and identify and extract local time and frequency features to generate a feature waveform set; Based on the characteristic waveform set, a support vector machine is used to analyze the characteristics, compare the real-time waveform and the historical waveform, identify the signal that deviates from the normal pattern, predict the potential fault, and generate a fault prediction result; Based on the fault prediction results and combined with the historical performance data of the obstruction lights, a performance model is constructed using a random forest algorithm to analyze the performance degradation trend and failure probability, and to predict the remaining life cycle of the obstruction lights to generate a life cycle assessment; Based on the life cycle assessment, a Kalman filter is applied to fuse data from multiple sources to establish a data view, and a multi-level decision tree is used to analyze the fused data, analyze the data structure and key features, and generate decision support information; Based on the decision support information, a self-organizing map network is used to perform topological mapping and clustering of the data, and a U-matrix visualization technique is used to map the data features into a two-dimensional space, identify data patterns and anomalies, and generate an anomaly pattern diagram; Based on the abnormal pattern diagram and in combination with the decision support information, a hierarchical analysis process is adopted to analyze monitoring performance and reliability, and based on the analysis results, maintenance, fault response and resource allocation solutions are provided to generate resource optimization and maintenance plans.

2. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: The signal purification data includes a signal sequence processed by Kalman filtering, a standard indicator of signal strength, and a signal frequency calibration result; the characteristic waveform set includes local time-frequency characteristics, frequency domain characteristics, and peak and trough information of the signal amplitude; the fault prediction result includes abnormal waveform identification records, classification information of predicted fault types, and an estimated value of the probability of fault occurrence; the life cycle assessment includes a degradation curve of the obstacle light performance, a predicted maintenance time point, and an estimated value of the remaining operating time; the decision support information includes a data view, key indicator analysis results, and potential risk point prompts; the abnormal pattern diagram includes a data cluster distribution diagram, an abnormal pattern marking area, and a correlation measurement of the clustering results; the resource optimization and maintenance plan includes a scheduled maintenance schedule, a resource reconfiguration plan, and a fault response priority ranking.

3. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on external sensors, electromagnetic signals emitted by aviation obstruction lights are collected. Signal processing algorithms are used to process the collected raw signals, eliminate noise interference, and standardize the signal format. The specific steps for generating signal purification data are as follows: Based on external sensors, the original electromagnetic signal is collected and denoised using a wavelet noise reduction algorithm. The denoised signal is then subjected to time series analysis to remove environmental noise and electromagnetic interference, generating a denoised signal. Based on the denoised signal, Z-score normalization is applied to normalize the signal amplitude and frequency, and amplitude-frequency analysis is performed on the normalized signal to determine the consistency and comparability of the signal to generate a standardized signal; Based on the standardized signal, a linear detrending algorithm is used to perform trend elimination processing on the signal, and a periodic analysis is performed on the detrended signal to eliminate non-stationary trends and optimize signal stability to generate a detrended signal; Based on the detrended signal, a time series synchronization algorithm is used to perform timestamp calibration on the signal, a data format conversion tool is used to perform data encoding conversion on the signal, and the signal format is matched to the analysis and processing requirements to generate signal purification data.

4. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the signal purification data, wavelet transform and Fourier transform are used to perform time-frequency analysis on the signal to extract key waveform features. The signal is decomposed through multiple transformations, and local time and frequency features are identified and extracted. The steps of generating a feature waveform set are as follows: Based on the signal purification data, a discrete wavelet transform is used to perform a multi-scale time-frequency analysis on the signal, and a local feature extraction is performed on the analysis result to identify the local time-frequency features of the signal and generate wavelet feature data; Based on the wavelet feature data, a fast Fourier transform is used to perform frequency domain analysis on the signal, and a global feature extraction is performed on the analysis result to identify the global frequency component of the signal and generate frequency domain feature data; Based on the wavelet feature data and the frequency domain feature data, feature cascade fusion is applied to fuse the local time-frequency features and the global frequency features, and information optimization is performed on the fused features to retain key feature information and generate comprehensive feature data; Based on the comprehensive feature data, principal component analysis is used to perform feature dimensionality reduction and optimization processing, and the information content of the features after dimensionality reduction and optimization is evaluated to screen the optimal features and generate a feature waveform set.

5. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the characteristic waveform set, a support vector machine is used to analyze the features, compare the real-time waveform with the historical waveform, identify signals that deviate from the normal pattern, and predict potential faults. The specific steps for generating the fault prediction result are as follows: Based on the characteristic waveform set, a support vector machine is used to analyze the characteristic data and perform pattern recognition. By comparing the real-time waveform with the historical waveform data, the signal characteristics that deviate from the normal waveform pattern are identified to generate an abnormal signal recognition result; Based on the abnormal signal recognition results, a K-means clustering algorithm is applied to perform pattern grouping processing on the abnormal signals, identify and distinguish different types of signal abnormal patterns, and generate a fault mode classification result; Based on the fault mode classification results, an Apriori algorithm is used to perform correlation analysis on the classified fault modes, identify potential correlations between differentiated fault modes and frequently occurring pattern combinations, and generate fault mode analysis results; Based on the failure mode analysis results, a Bayesian estimation method is used to perform probability analysis on the identified failure modes and perform failure probability estimation, calculate and predict the occurrence probabilities of multiple failure modes, and generate failure prediction results.

6. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the fault prediction results and combined with the historical performance data of the obstruction lights, a performance model is constructed using the random forest algorithm to analyze the performance degradation trend and failure probability, and to predict the remaining life cycle of the obstruction lights. The specific steps for generating a life cycle assessment are as follows: Based on the fault prediction results and combined with the historical performance data of the obstacle lights, a random forest algorithm is used to analyze the performance trend of the data set, identify performance degradation trends and key points, and generate performance degradation analysis results; Based on the performance degradation analysis results, an autoregressive integrated moving average model is used to perform time series analysis on the performance data, and a prediction model for obstacle light performance changes is established to generate a performance change model; Based on the performance change model, the Cox proportional risk model is used to analyze the life of the obstruction light, perform risk and life prediction, estimate the remaining life cycle and failure probability of the obstruction light, and generate life prediction analysis results; Based on the life prediction analysis results, multi-criteria decision analysis is used to analyze the performance and remaining life of the obstacle lights, and the performance and life are evaluated to determine the status and maintenance requirements of the obstacle lights and generate a life cycle assessment.

7. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the life cycle assessment, the Kalman filter is applied to fuse data from multiple sources to establish a data view. The fused data is analyzed using a multi-level decision tree, and the data structure and key features are analyzed to generate decision support information. The specific steps are as follows: Based on the life cycle assessment, a Kalman filter is used to perform real-time updating and noise elimination on multi-source data, and a decision-layer fusion method is used to process data provided by multiple sensors, perform data accuracy and uncertainty optimization operations, and generate an optimized data view; Based on the optimized data view, random forest and decision tree algorithms are applied to analyze the data set, voting on the results is performed by constructing multiple decision trees, and feature criticality assessment and data classification are performed to generate decision tree analysis results; Based on the decision tree analysis results, principal component analysis is used to perform dimensionality reduction processing, optimize the number of features by converting the data into a new feature space, adjust the feature correlation and compress the data to generate a structural analysis result; Based on the structural analysis results, an information integration framework and data warehouse technology are used to extract data and refine information by merging key information and features from multiple analysis stages to generate decision support information.

8. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the decision support information, a self-organizing map network is used to perform topological mapping and clustering of the data. The data features are mapped to a two-dimensional space through U-matrix visualization technology, and data patterns and anomalies are identified. The specific steps for generating an anomaly pattern map are as follows: Based on the decision support information, a self-organizing map network is used to perform topological mapping and feature clustering processing on the data, and pattern recognition and classification are performed through network self-learning adjustment to generate a topological mapping clustering result; Based on the topological mapping clustering results, the U matrix visualization technology is used to perform two-dimensional spatial mapping processing on the data features, the data point positions are represented by color grayscale, and the data patterns are visualized to generate a data feature visualization graph; Based on the data feature visualization graph, hierarchical density clustering is used to mine the data, and by detecting dense areas of the data, abnormal pattern recognition and normal pattern division are performed to generate a pattern recognition graph; Based on the pattern recognition graph, a clustering quality assessment method is used to perform quality inspection on the identified pattern, and clustering indicators are calculated to perform pattern assessment and quality judgment to generate an abnormal pattern graph.

9. The remote monitoring method for aviation obstruction lights according to claim 1, characterized in that: Based on the abnormal pattern diagram and the decision support information, a hierarchical analysis process is used to analyze monitoring performance and reliability, and based on the analysis results, maintenance, fault response and resource allocation plans are provided. The specific steps for generating a resource optimization and maintenance plan are as follows: Based on the abnormal pattern diagram and in combination with the decision support information, a hierarchical analysis process is used to analyze the potential impact of the failure mode on the performance, evaluate the monitoring performance and reliability, and perform failure cause analysis and impact assessment to generate a performance reliability assessment result; Based on the performance reliability assessment results, a risk priority number scoring method is applied to calculate a risk priority number, including the product of severity, probability of occurrence, and difficulty of detection, to perform quantitative assessment and ranking of risks and generate a risk priority assessment result; Based on the risk priority assessment results, a linear programming model is used to establish a linear programming problem to balance resource allocation and cost, optimize the monitoring resource configuration, and generate resource optimization configuration results; Based on the resource optimization configuration results, particle swarm optimization is adopted to combine resource configuration and risk assessment results to establish a maintenance plan and fault response strategy, and to perform resource allocation to generate a resource optimization and maintenance plan.

10. A remote monitoring system for aviation obstruction lights, characterized in that: According to any one of claims 1 to 9, the remote monitoring method for aviation obstruction lights comprises a signal processing module, a feature extraction module, a fault prediction module, a life cycle assessment module, a decision support module, and a resource optimization module; The signal processing module collects raw electromagnetic signals based on external sensors, uses a wavelet noise reduction algorithm to denoise the signals, applies Z-score standardization to normalize the signal amplitude and frequency, uses a linear detrending algorithm to eliminate the trend of the signals, uses a time series synchronization algorithm to perform time stamp calibration on the signals, performs signal data encoding conversion, and generates a purified signal data set; The feature extraction module uses discrete wavelet transform to perform multi-scale time-frequency analysis on the signal based on the purified signal data set, uses fast Fourier transform to perform frequency domain analysis on the signal and extract global features, applies feature cascade fusion to fuse local time-frequency features and global frequency features, and then performs feature dimensionality reduction and optimization through principal component analysis to generate a feature waveform set; The fault prediction module uses a support vector machine to perform pattern recognition on the feature data based on the feature waveform set, identifies signal features that deviate from the normal waveform pattern, applies the K-means clustering algorithm to perform pattern grouping processing on abnormal signals, uses the Apriori algorithm to perform association analysis on the classified fault patterns, and uses the Bayesian estimation method to perform probability analysis on the identified fault patterns to generate fault prediction analysis results; The life cycle assessment module is based on the results of the fault prediction analysis and combines the historical performance data of the obstruction lights. It uses the random forest algorithm to analyze the performance trend of the data set, adopts the autoregressive integral moving average model to establish a prediction model for the performance change of the obstruction lights, and uses the Cox proportional hazard model to predict the risk and life of the obstruction lights. Then, through multi-criteria decision analysis, the performance and remaining life of the obstruction lights are evaluated to generate the life cycle assessment results. The decision support module uses a Kalman filter to perform real-time updates and noise elimination on multi-source data based on the results of life cycle assessments, applies random forest and decision tree algorithms to analyze data sets and assess feature criticality, performs data dimensionality reduction through principal component analysis, and then uses an information integration framework and data warehouse technology to extract and refine key information and features from multiple analysis stages to generate a decision support information set. The resource optimization module is based on the decision support information set and uses a self-organizing map network to perform topological mapping and feature clustering on the data. Through U-matrix visualization technology, it performs two-dimensional spatial mapping on the data features. Hierarchical density clustering is used to mine and recognize patterns on the data. Combined with the performance evaluation results, the particle swarm optimization algorithm is used to establish maintenance plans and fault response strategies to generate resource optimization and maintenance strategies.

Images