Park facility operation state monitoring method and system based on multi-mode perception

Abstract

The invention relates to the related technical field of multi-modal sensing monitoring, in particular to a park facility operation state monitoring method and system based on multi-modal sensing, and the method comprises the steps: collecting multi-modal sensing data, generating a fusion matrix, extracting operation features, constructing a facility state evolution graph, configuring a monitoring period, and recognizing a potential fault mode. And matching with historical data to optimize a monitoring rule, and outputting a monitoring report, thereby solving the technical problems that cross-modal associated potential fault features cannot be effectively extracted, and the accuracy and comprehensiveness of fault identification are affected, and realizing construction of a facility state evolution graph, time prediction of a health degree index and space labeling of a fault risk probability. The method has the technical effects that the hidden faults are recognized in advance, meanwhile, the sensitivity of fault feature extraction is improved through multi-modal data backtracking analysis and state transition feedback optimization, the adaptive correlation model is continuously optimized by adopting incremental learning and model iteration, and the accuracy and comprehensiveness of fault recognition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to multimodal sensing monitoring, and in particular to a method and system for monitoring the operating status of park facilities based on multimodal sensing. Background Art

[0002] With the continuous development of technologies such as the Internet of Things, big data, and artificial intelligence, the technology for monitoring the operating status of park facilities is also constantly improving. Smart parks have integrated advanced information technologies such as GIS geographic information systems, Internet of Things technology, BIM technology, 5G technology, cloud computing technology, big data technology, and artificial intelligence technology to achieve intelligent transformation and upgrading of infrastructure, public safety, energy management, environmental protection, etc. within the park, providing more comprehensive and accurate data support for monitoring the operating status of park facilities.

[0003] However, existing fault identification rules are mostly based on static thresholds or single indicators (such as health index), and are unable to adaptively adjust the monitoring cycle and threshold range according to dynamic parameters such as ambient temperature and load rate, resulting in false alarms or missed alarms. Therefore, how to effectively apply multimodal sensing technology to the operation status monitoring of park facilities and improve the accuracy of monitoring has become a problem that needs to be solved urgently.

[0004] In summary, the existing technology has the technical problem that the pattern matching mechanism of the historical fault database is single and cannot effectively extract potential fault features associated across modalities, thus affecting the accuracy and comprehensiveness of fault identification. Summary of the invention

[0005] This application provides a campus facility operation status monitoring system based on multimodal perception, aiming to solve the technical problem that the historical fault database in the prior art has a single pattern matching mechanism and cannot effectively extract potential fault features associated across modalities, thus affecting the accuracy and comprehensiveness of fault identification.

[0006] In view of the above problems, the technical solution to implement this application is: On the one hand, the present application provides a method for monitoring the operating status of park facilities based on multimodal perception, wherein the method includes: collecting multimodal perception data of each park facility in the target park, generating a fusion data matrix through alignment processing, and inputting it into a pre-trained adaptive association model to extract facility operation characteristics; based on the facility operation characteristics, constructing a facility status evolution map, including a facility health index, an energy consumption abnormality coefficient and a failure risk probability; configuring a first monitoring period according to the facility health index and the energy consumption abnormality coefficient in the facility status evolution map, and identifying a class of potential failure modes under an initial fault identification rule; configuring a second monitoring period according to the failure risk probability and the energy consumption abnormality coefficient in the facility status evolution map, and identifying two classes of potential failure modes under an initial fault identification rule; performing pattern matching on the first monitoring period and the first class of potential failure modes, the second monitoring period and the second class of potential failure modes, and a historical fault database, so as to optimize the monitoring period and fault identification rules of each park facility with the matching results, and output a fault monitoring report.

[0007] On the other hand, the present application provides a park facility operation status monitoring system based on multimodal perception, wherein the system includes: a data acquisition module, which is used to collect multimodal perception data of each park facility in the target park, generate a fusion data matrix through alignment processing, and input it into a pre-trained adaptive association model to extract facility operation characteristics; a graph construction module, which is used to construct a facility state evolution graph based on the facility operation characteristics, including a facility health index, an energy consumption abnormality coefficient and a fault risk probability; a first identification module, which is used to configure a first monitoring period according to the facility health index and the energy consumption abnormality coefficient in the facility state evolution graph, and identify a class of potential fault modes under the initial fault identification rule; a second identification module, which is used to configure a second monitoring period according to the fault risk probability and the energy consumption abnormality coefficient in the facility state evolution graph, and identify two classes of potential fault modes under the initial fault identification rule; a pattern matching module, which is used to pattern match the first monitoring period and the first class of potential fault modes, the second monitoring period and the second class of potential fault modes with the historical fault database, optimize the monitoring period and fault identification rules of each park facility with the matching results, and output a fault monitoring report.

[0008] In summary, one or more technical solutions provided in this application realize the construction of a facility status evolution map, combining the time prediction of the health index and the spatial annotation of the fault risk probability to identify hidden faults in advance. At the same time, through multimodal data backtracking analysis and state transfer feedback optimization, the sensitivity of fault feature extraction is improved, and incremental learning and model iteration are used to continuously optimize the adaptive correlation model to improve the accuracy and comprehensiveness of fault identification. Technical effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flow chart of a method for monitoring the operation status of park facilities based on multimodal perception is provided for this application; Figure 2 A structural diagram of a park facility operation status monitoring system based on multimodal perception is provided for this application.

[0010] Explanation of the reference numerals: data acquisition module M100, map construction module M200, first recognition module M300, second recognition module M400, pattern matching module M500. DETAILED DESCRIPTION

[0011] Embodiment 1 The present application is described in detail below with reference to the accompanying drawings. Figure 1 As shown, the present application provides a method for monitoring the operation status of park facilities based on multimodal perception, wherein the method includes: S1: Collect multimodal perception data of each park facility in the target park, generate a fusion data matrix through alignment processing, and input it into the pre-trained adaptive association model to extract the facility operation characteristics; S2: Based on the facility operation characteristics, construct a facility state evolution map, including facility health index, energy consumption abnormality coefficient and failure risk probability.

[0012] Specifically, multimodal perception data refers to data related to the operating status of campus facilities collected through a variety of different types of sensors, including operating voiceprint signals (sound features collected by acoustic sensors), infrared thermal imaging maps (temperature distribution images collected by infrared thermal imaging equipment), and current harmonic features (harmonic components in current collected by current sensors). Since the data collection time, frequency, and method of different modalities may differ, they are synchronized through time axis alignment operations to ensure temporal consistency. For example, the sampling time windows of the operating voiceprint signals, infrared thermal imaging maps, and current harmonic features are aligned to eliminate phase delay errors. The aligned multimodal data are integrated into a unified data structure to form a fused data matrix containing feature information from different modalities.

[0013] The pre-trained adaptive association model is pre-trained using a large amount of historical data. The pre-trained adaptive association model can dynamically learn the cross-modal correlation between multimodal data, that is, the relationship between different modal data. For example, it can learn the correlation between the anomaly in the operating voiceprint signal and the current harmonic characteristics, so as to more accurately extract the facility operation characteristics; the facility operation characteristics refer to the key information extracted from the fused data matrix that can reflect the operating status of the facility, including the vibration frequency, temperature change rate, amplitude of current harmonics, etc. of the equipment, which is the basis for constructing the facility state evolution map; the facility state evolution map includes the facility health index (a quantitative indicator reflecting the overall health status of the equipment), the energy consumption anomaly coefficient (an indicator to measure whether the energy consumption of the equipment is abnormal) and the failure risk probability (predicting the possibility of equipment failure). The facility state evolution map is presented in the form of time series analysis and spatial topology annotation, which can intuitively display the operating status of the facility and its changing trend.

[0014] First, multimodal perception data of each facility in the target park is collected, including operation voiceprint signals, infrared thermal imaging maps and current harmonic features, and synchronously processed to eliminate the phase delay error between different modal data. The data are mapped into the feature space to generate a fused data matrix. Next, the fused data matrix is ​​input into the pre-trained adaptive association model to extract the facility operation characteristics and construct a facility state evolution map, including the facility health index, energy consumption abnormality coefficient and failure risk probability.

[0015] By collecting multimodal perception data and aligning it, the operating status of the facility can be fully and accurately reflected. The pre-trained adaptive association model can dynamically learn the correlation between multimodal data and extract more representative operating characteristics. The facility status evolution map provides a basis for subsequent monitoring cycle configuration and fault mode identification. For example, the facility health index can be used to evaluate the overall health status of the equipment, the energy consumption abnormality coefficient can timely detect the energy consumption abnormality of the equipment, and the failure risk probability can provide early warning of potential failure risks, so that the monitoring system can grasp the operating status of the facility more comprehensively and accurately, and provide strong support for subsequent fault diagnosis and optimization.

[0016] S3: According to the facility health index and energy consumption anomaly coefficient in the facility status evolution map, configure the first monitoring cycle, and identify a type of potential failure mode under the initial fault identification rule; S4: According to the failure risk probability and energy consumption anomaly coefficient in the facility status evolution map, configure the second monitoring cycle, and identify two types of potential failure modes under the initial fault identification rule; S5: Pattern match the first monitoring cycle and one type of potential failure mode, the second monitoring cycle and two types of potential failure modes with the historical fault database, optimize the monitoring cycle and fault identification rules of each park facility with the matching results, and output a fault monitoring report.

[0017] Specifically, the monitoring cycle refers to the time interval for monitoring the operating status of campus facilities. The monitoring cycle can be dynamically adjusted according to the health status and failure risk of the facilities. For example, for campus facilities with higher failure risks, the monitoring cycle will be shortened to obtain data more frequently and detect potential problems in a timely manner. The initial fault identification rules are used to preliminarily determine whether there are potential faults in the facilities. They are usually formulated based on the normal operating characteristics and historical fault data of the facilities. For example, when the facility health index is lower than a certain threshold or the energy consumption abnormality coefficient exceeds a certain range, fault identification may be triggered.

[0018] Category I potential failure modes refer to the failure types identified through the first monitoring cycle and initial failure identification rules, which are usually closely related to the health index and energy consumption abnormality coefficient of the facility, including mechanical wear, circuit overload, etc., and are a direct reflection of the current operating status of the facility; Category II potential failure modes refer to the failure types identified through the second monitoring cycle and initial failure identification rules, which are usually related to the failure risk probability and energy consumption abnormality coefficient, including heat dissipation abnormality, harmonic distortion, etc., and pay more attention to the potential risks of the facility and possible problems in the future; Pattern matching refers to comparing and analyzing the currently identified Category I potential failure modes and Category II potential failure modes with the known failure modes in the historical failure database to determine the similarity between Category I potential failure modes and Category II potential failure modes; Through pattern matching, the accuracy of the current failure mode can be verified, and a basis can be provided for the optimization of subsequent monitoring cycles and failure identification rules; The fault monitoring report summarizes the operating status of the facility, the identified potential failure modes, the optimized monitoring cycles and failure identification rules, etc., to provide support for the maintenance and management of park facilities.

[0019] According to the facility health index and energy consumption anomaly coefficient in the facility status evolution map, the first monitoring cycle is dynamically configured. For example, when the facility health index is low or the energy consumption anomaly coefficient is high, the monitoring cycle is shortened to monitor the facility status more frequently. Under the initial fault identification rules, a type of potential failure mode is identified through real-time monitoring data, such as mechanical wear or circuit overload; according to the failure risk probability and energy consumption anomaly coefficient in the facility status evolution map, the second monitoring cycle is dynamically configured. The second monitoring cycle pays more attention to the potential risks of the facility, and identifies the second type of potential failure mode through the initial fault identification rules, such as heat dissipation abnormality or harmonic distortion.

[0020] The first monitoring cycle and the first type of potential failure mode, the second monitoring cycle and the second type of potential failure mode are pattern matched with the historical fault database. Through pattern matching, the accuracy of the current failure mode is verified, and the monitoring cycle and fault identification rules are optimized according to the matching results. For example, if the current failure mode is highly similar to a failure mode in the historical database, the monitoring cycle can be adjusted to track the development of the failure mode more accurately. A fault monitoring report is generated to summarize the operating status of the facility, the identified potential failure modes, the optimized monitoring cycle and fault identification rules, etc., to provide decision support for the maintenance and management of park facilities.

[0021] By dynamically configuring the monitoring cycle and identifying potential failure modes, the system can flexibly adjust the monitoring strategy according to the actual operating status of the facility, thereby improving the accuracy and efficiency of fault identification. For example, the identification of the first type of potential failure mode focuses on the direct reflection of the current operating status, while the identification of the second type of potential failure mode focuses more on the risks that may arise in the future. Through pattern matching, the system can use historical data to optimize the monitoring cycle and fault identification rules, forming a closed loop of "perception-diagnosis-optimization", improving the intelligence level of the system, and the output fault monitoring report provides comprehensive and accurate information support for the maintenance and management of park facilities, ensuring the safe operation of the facilities.

[0022] Furthermore, the application method includes: The multimodal sensing data includes operating voiceprint signals, infrared thermal imaging maps and current harmonic characteristics; non-invasive current sensors are deployed in the power supply circuits corresponding to each park facility in the target park to obtain current harmonic characteristics including odd harmonic amplitudes and phase offsets; the surface temperature field of the park facilities is obtained to generate infrared thermal imaging maps with timestamps; and a distributed acoustic sensor array is used to collect operating voiceprint signals.

[0023] Specifically, multimodal perception data refers to data collected through a variety of different types of sensors, which are used to comprehensively monitor the operating status of park facilities. Multimodal perception data includes operating voiceprint signals, infrared thermal imaging maps and current harmonic characteristics, which reflect the operating status of facilities from different angles and can provide more comprehensive information; operating voiceprint signals refer to the sound signals generated by the equipment during operation, which are collected by distributed acoustic sensor arrays. They reflect the mechanical vibration, abnormal noise and other information of the equipment, and are used to determine whether the equipment has mechanical failures or wear; infrared thermal imaging maps refer to the surface temperature distribution images of the facilities obtained by infrared thermal imaging equipment, which reflect the thermal distribution of the equipment and help identify overheating areas, thereby determining whether there are heat dissipation problems or potential failure points. The maps are timestamped and are used to record the time series of temperature changes.

[0024] The harmonic characteristics of current refer to the harmonic components in the current collected by non-invasive current sensors. Harmonics are components of non-fundamental frequencies in current, which are usually caused by electrical failures, aging or non-linear loads of equipment. The harmonic characteristics of current include the amplitude and phase offset of odd harmonics, which reflect the electrical health status of the equipment. Non-invasive current sensors are installed in the power supply circuit. They can measure current without direct contact with the conductor. They obtain current information by sensing the magnetic field generated by the current. They are easy to install, safe and reliable. Distributed acoustic sensor array refers to a group of acoustic sensors distributed around the park facilities, which are used to collect sound signals when the equipment is running. Through the coordinated work of multiple sensors, the sound source can be located more accurately and the sound characteristics can be analyzed.

[0025] Deploy non-invasive current sensors in the power supply circuits of various facilities in the target park to collect current harmonic characteristics; install distributed acoustic sensor arrays around the facilities to collect operating voiceprint signals; use infrared thermal imaging equipment to obtain the temperature field on the surface of the facilities and generate infrared thermal imaging maps with timestamps; preferably, the current harmonic characteristics refer to the real-time collection of odd harmonic amplitudes and phase offsets in the current by non-invasive current sensors, which reflect the electrical health status of the equipment. For example, an abnormal increase in the harmonic amplitude may indicate that the equipment has an electrical fault; the infrared thermal imaging map refers to the infrared thermal imaging equipment regularly scanning the surface of the facility, generating a temperature distribution map, and recording a timestamp. By analyzing the temperature changes, the heat dissipation problem of the equipment or the local overheating area can be identified; the operating voiceprint signal refers to the sound signal collected by the distributed acoustic sensor array when the equipment is running, and the voiceprint characteristics are extracted by acoustic analysis technology. For example, abnormal high-frequency noise may indicate that the equipment has mechanical wear.

[0026] By deploying a variety of sensors to collect multimodal perception data, comprehensive and accurate information support is provided for subsequent data fusion and fault diagnosis. The operating soundprint signal, infrared thermal imaging spectrum and current harmonic characteristics reflect the operating status of the facility from the mechanical, thermal and electrical dimensions respectively, making up for the shortcomings of single-modal monitoring. For example, mechanical failure may cause abnormal operating soundprint signals, electrical failure may cause current harmonic changes, and heat dissipation problems will be reflected in infrared thermal imaging maps as local overheating. Through the coordinated monitoring of multimodal data, potential faults can be identified more comprehensively and the accuracy and reliability of the monitoring system can be improved.

[0027] Furthermore, the multimodal sensing data of each park facility in the target park is collected to generate a fusion data matrix through alignment processing. The present application method includes: According to the operation timeline of the park facilities, the sampling time windows of the operation voiceprint signal, infrared thermal imaging spectrum and current harmonic characteristics are aligned; based on the aligned sampling time windows, the phase delay error between the multimodal data is eliminated, and the data is mapped into the feature space to generate the fused data matrix.

[0028] Specifically, the operation timeline refers to the time series during the operation of the park facilities, which is used to record and synchronize the collection time points of different modal data. The operation timeline is the basis for multimodal data alignment, ensuring that the data collected by different sensors can correspond in time; the sampling time window refers to the collection time range set for each modal data on the operation timeline. Since the collection frequency and method of the operation voiceprint signal, infrared thermal imaging spectrum and current harmonic characteristics may be different, alignment operations need to be performed through the sampling time window for subsequent fusion processing; phase delay error refers to the deviation in time between different modal data. For example, the operation voiceprint signal and current harmonic characteristics may be inconsistent in time due to differences in sensor response time or data transmission delays, affecting the accuracy of data fusion.

[0029] The feature space is used to map data of different modalities into a unified and processable dimension. Each dimension in the feature space represents a feature, such as the frequency characteristics of running voiceprint signals, the temperature gradient characteristics of infrared thermal imaging maps, or the amplitude characteristics of current harmonics. By mapping the data into the feature space, data fusion and analysis can be conveniently performed; the aligned multimodal data are integrated into a matrix structure. Each row of the fused data matrix represents the multimodal features of a time point, and each column of the fused data matrix represents the features of a modality. The fused data matrix is ​​the basis for subsequent feature extraction and model training.

[0030] According to the operation timeline of the park facilities, the sampling time windows of the operation voiceprint signal, infrared thermal imaging spectrum and current harmonic characteristics are aligned. Specifically, a reference mode (such as infrared thermal imaging spectrum) is selected, and then the sampling time windows of other modal data are adjusted to make them consistent with the time points of the reference mode. For example, if the sampling frequency of the operation voiceprint signal is 100Hz, the sampling frequency of the infrared thermal imaging spectrum is 1Hz, and the sampling frequency of the current harmonic characteristics is 50Hz, it is necessary to align the time points through interpolation; within the aligned sampling time window, eliminate the phase delay error between different modal data. For example, a time delay estimation algorithm (such as the cross-correlation method) can be used to calculate the time delay between the operation voiceprint signal and the current harmonic characteristics, and the phase delay error can be eliminated by adjusting the time series of the data.

[0031] Features are extracted from the aligned multimodal data and mapped to the feature space. For example, frequency features and energy features are extracted from the operating voiceprint signal, temperature gradients and local temperature change rates are extracted from infrared thermal imaging maps, and harmonic amplitudes and phase offsets are extracted from current harmonic features. Then, they are integrated into a high-dimensional feature space to form a fused data matrix. By aligning the sampling time windows and eliminating phase delay errors, the temporal consistency and accuracy of the multimodal data are ensured. Mapping the data to the feature space further improves the processability and analysis efficiency of the data. The fused data matrix provides a high-quality data foundation for subsequent feature extraction, model training and fault diagnosis. For example, by eliminating phase delay errors, misjudgments due to time inconsistencies can be avoided. Through the integration of feature spaces, the advantages of multimodal data can be fully utilized to improve the sensitivity and accuracy of fault identification.

[0032] Furthermore, based on the facility operation characteristics, a facility state evolution map is constructed, and the method of the present application includes: Extracting spatial correlation features in the fused data matrix; predicting the time evolution trend of the facility health index based on the spatial correlation features in the fused data matrix; determining the failure risk probability through the prediction results, and marking it on the topological node of the facility state evolution map.

[0033] Specifically, spatial correlation features refer to the correlation in spatial distribution between different modal data in the fused data matrix. For example, the current harmonic features of the equipment may be spatially correlated with local temperature changes (infrared thermal imaging maps) or mechanical vibrations (operating soundprint signals), reflecting the interaction between different components inside the equipment; the time evolution trend refers to the changing trend of the facility health index over time. By analyzing the spatial correlation features in the fused data matrix, we can predict the future changes in the facility health index, discover potential health problems in advance, and provide a basis for preventive maintenance.

[0034] The failure risk probability refers to the possibility of a device failing at a certain point in the future. It is calculated based on the time evolution trend of the facility health index and the current operating status. The determination of the failure risk probability provides a risk warning function for the facility state evolution map. The topological node of the facility state evolution map refers to the key position in the facility state evolution map, which is used to mark the health status, energy consumption anomalies, failure risk and other information of the equipment. The topological node can be a specific component of the equipment or a key point of the operating status. The failure risk probability is marked on the topological node of the facility state evolution map to intuitively display the health status and potential risks of the equipment.

[0035] Extract spatial correlation features between different modal data from the fused data matrix, for example, by analyzing the correlation between current harmonic features and local temperature changes in infrared thermal imaging maps, or the correlation between operating voiceprint signals and equipment surface temperature; based on the extracted spatial correlation features, use time series analysis methods (such as ARIMA model, long short-term memory network LSTM) to predict future changes in the facility health index. For example, if the current harmonic amplitude of the equipment is positively correlated with the local temperature change, and the correlation increases over time, it indicates that the health index is declining.

[0036] According to the time evolution trend of the health index, combined with the current energy consumption anomaly coefficient and other operating status indicators, the fault risk probability is calculated. For example, if the health index shows a downward trend and the energy consumption anomaly coefficient is high, the fault risk probability increases, and the calculated fault risk probability is marked on the topological node of the facility state evolution map; by extracting spatial correlation features, we can deeply understand the intrinsic connection between different modal data, so as to more accurately evaluate the health status of the equipment, and predict the time evolution trend of the facility health index, which provides early warning for preventive maintenance and avoids sudden failures. The determination and marking of the fault risk probability further enhances the function of the facility state evolution map, enabling it to intuitively display the health status and potential risks of the equipment, and provides strong support for the intelligence and automation of the entire monitoring system.

[0037] Furthermore, according to the facility health index and the energy consumption anomaly coefficient in the facility state evolution map, a first monitoring period is configured, and a class of potential failure modes is identified under the initial failure identification rule. The method of the present application also includes: A first threshold range corresponding to the facility health index and the energy consumption anomaly coefficient is configured, and the first threshold range is adaptively adjusted with the ambient temperature and the load rate; in the initial fault identification rule, the first threshold range and the first monitoring period are used to establish a first state assessment decision chain corresponding to a type of potential fault mode, and the facility operation data is monitored in real time; the first state assessment decision chain is used to perform multimodal data backtracking analysis to extract multiple abnormal feature fragments of the fault; and the first state assessment decision chain is optimized for state transfer feedback based on the multiple abnormal feature fragments.

[0038] Specifically, the first threshold range refers to the initial threshold interval set for the facility health index and the energy consumption abnormality coefficient, which is used to determine whether the facility operation status is normal. The threshold range will be dynamically adjusted according to the ambient temperature and load rate to adapt to the operating status under different working conditions; adaptive adjustment means that the threshold range will change dynamically according to real-time environmental conditions (such as temperature, load rate, etc.). For example, when the ambient temperature rises, the threshold of the energy consumption abnormality coefficient may be relaxed accordingly, because the energy consumption of the equipment may naturally increase when operating in a high temperature environment.

[0039] The first state assessment decision is used to assess the operating state of the facility based on the first threshold range and the first monitoring cycle; multimodal data retrospective analysis refers to retrospective analysis of historical multimodal data after a potential fault is detected, and extraction of abnormal feature fragments before and after the fault occurs, including abnormal noise in the operating voiceprint signal, abnormal temperature areas in infrared thermal imaging maps, or mutations in current harmonic features; state transfer feedback optimization refers to the optimization of the first state assessment decision chain based on the results of multimodal data retrospective analysis, by adjusting the decision logic or threshold range, so that it can more accurately identify fault modes and reduce false alarms or omissions.

[0040] According to the normal operating status of the facility, the initial threshold range of the health index and the energy consumption abnormality coefficient is set. At the same time, the ambient temperature and load rate are introduced as dynamic adjustment factors so that the threshold range can adaptively change according to the real-time working conditions. For example, when the load rate increases, the threshold of the energy consumption abnormality coefficient may be appropriately increased to avoid misjudgment as an abnormality due to an increase in normal load. Combined with the first threshold range and the first monitoring cycle, a decision logic for real-time monitoring of the operating status of the facility is constructed. For example, if the facility health index is lower than the set threshold or the energy consumption abnormality coefficient exceeds the threshold, an alarm for a potential failure mode is triggered.

[0041] When a class of potential fault modes is identified, a retrospective analysis is performed on the historical multimodal data to extract multiple abnormal feature fragments before and after the fault occurs; based on the results of the retrospective analysis, the first state evaluation decision chain is optimized. For example, if a certain abnormal feature fragment is found to appear frequently in multiple faults, it can be added to the decision logic to enhance the accuracy of the decision chain; or the threshold range is adjusted according to the actual fault situation to make it more in line with the actual operating state; by adaptively adjusting the threshold range, it can more flexibly respond to the operating state under different working conditions and reduce false alarms or omissions caused by environmental changes. The first state evaluation decision chain provides a clear logical basis for real-time monitoring and can quickly identify a class of potential fault modes. Multimodal data retrospective analysis and state transfer feedback optimization further enhance the intelligence level of the system, enabling it to dynamically adjust the decision logic according to actual operating data and improve the accuracy and reliability of fault identification.

[0042] Furthermore, the first monitoring period and the first type of potential failure mode, the second monitoring period and the second type of potential failure mode are pattern matched with the historical failure database. The method of the present application includes: Configure a second threshold range corresponding to the facility health index and the energy consumption anomaly coefficient; in the initial fault identification rule, use the second threshold range and the second monitoring period to establish a second state evaluation decision chain corresponding to the second type of potential failure mode; use the first state evaluation decision chain and the second state evaluation decision chain to configure a bidirectional traversal architecture, and the bidirectional traversal architecture is used to cross-validate with one type of potential failure mode and two types of potential failure modes.

[0043] Specifically, similar to the first threshold range, the second threshold range is another set of threshold intervals set for the facility health index and the energy consumption abnormality coefficient, which is specifically used to identify the second type of potential failure mode. The second threshold range may be adjusted according to different failure characteristics or more stringent monitoring requirements to adapt to the identification of the second type of potential failure mode; the second state evaluation decision chain is a decision logic constructed based on the second threshold range and the second monitoring cycle, which is used to evaluate whether the facility has a second type of potential failure mode, which is similar to the first state evaluation decision chain.

[0044] The bidirectional traversal architecture is used to cross-validate the system architecture of the first and second potential failure modes. By running the first state evaluation decision chain and the second state evaluation decision chain at the same time and exchanging and verifying information between the two, the accuracy and comprehensiveness of fault identification are ensured. The bidirectional traversal architecture reduces the possibility of false alarms and missed alarms by comparing the identification results of the two types of failure modes. According to the characteristics of the second type of potential failure mode, a second threshold range is set for the facility health index and the energy consumption abnormality coefficient. In combination with the second threshold range and the second monitoring cycle, a decision logic for identifying the second type of potential failure mode is constructed. For example, if the energy consumption abnormality coefficient of the facility exceeds the second threshold range and the failure risk probability reaches a certain level, the alarm for the second type of potential failure mode is triggered.

[0045] The first-state evaluation decision chain and the second-state evaluation decision chain are integrated into the bidirectional traversal architecture. The bidirectional traversal architecture ensures the accuracy and comprehensiveness of fault identification by cross-validating the identification results of the first and second potential fault modes. For example, if the first-state evaluation decision chain identifies a first-class potential fault mode, and the second-state evaluation decision chain does not identify the corresponding second-class potential fault mode, further data analysis is required to determine whether there are false alarms or omissions. By configuring the second threshold range and the second-state evaluation decision chain, more accurate monitoring and identification of the second-class potential fault mode can be performed. The introduction of the bidirectional traversal architecture further improves the reliability of the system. By cross-validating the identification results of the first and second-class potential fault modes, the false alarm or omission problem that may be caused by a single decision chain is reduced. The double verification mechanism ensures the accuracy of fault identification and enhances the accuracy and comprehensiveness of fault identification.

[0046] Furthermore, the bidirectional traversal architecture is used to perform cross-validation with a first type of potential failure mode and a second type of potential failure mode. The method of the present application includes: When the first type of potential failure mode and the second type of potential failure mode are triggered simultaneously, the priority sorting mechanism of the bidirectional traversal architecture is activated; under the priority sorting mechanism, a virtual aggregate index item is added; and the virtual aggregate index item is used to perform pattern matching in a historical failure database.

[0047] Specifically, in a bidirectional traversal architecture, when a type I potential failure mode and a type II potential failure mode are triggered at the same time, the two failure modes need to be prioritized. The priority sorting mechanism is used to determine which failure mode is more urgent or more critical, so as to give priority to it. The priority is usually determined based on factors such as the severity of the failure, the probability of risk, and the scope of impact. A virtual aggregate index item refers to a temporary index item introduced in the pattern matching process to improve the matching efficiency and accuracy. The virtual aggregate index item is dynamically generated based on the characteristics of the currently identified failure mode, and is used to quickly locate historical records similar to the current failure mode in the historical failure database, which can accelerate the pattern matching process and improve the matching accuracy. Pattern matching refers to comparing and analyzing the currently identified failure mode (including type I and type II potential failure modes) with the records in the historical failure database to determine whether there are similar failure modes.

[0048] When a type 1 potential failure mode and a type 2 potential failure mode are triggered at the same time, the bidirectional traversal architecture activates the priority sorting mechanism. The system sorts the two failure modes according to preset rules (such as failure risk probability, health index, energy consumption abnormality coefficient, etc.) to determine which failure mode is more urgent or more critical. For example, if the risk probability of a type 1 potential failure mode is higher, it will be given a higher priority. Under the priority sorting mechanism, virtual aggregate index items are dynamically generated according to the characteristics of the current failure mode. They are generated based on the key characteristics of the failure mode (such as a specific health index range, energy consumption abnormality coefficient, failure risk probability, etc.) and are used to quickly locate similar failure records in the historical failure database.

[0049] Using virtual aggregate index items, pattern matching is performed in the historical fault database. The system will quickly retrieve historical records similar to the current fault mode based on the virtual aggregate index items and calculate the similarity. For example, by calculating the characteristic vector distance (such as Euclidean distance or cosine similarity) between the current fault mode and the historical record, it is determined whether there is a matching historical fault mode. The matching result will be used to verify the accuracy of the current fault mode. When the first and second potential fault modes are triggered at the same time, the priority sorting mechanism can quickly determine which fault mode is more urgent, so as to reasonably allocate resources and give priority to key issues. The introduction of virtual aggregate index items significantly improves the efficiency and accuracy of pattern matching, allowing the system to quickly find records similar to the current fault mode in massive historical fault data, and use the pattern matching mechanism to verify the accuracy of the current fault mode, providing strong support for subsequent fault diagnosis and optimization, and further improving the intelligence level and reliability of the system.

[0050] Furthermore, the virtual aggregate index item is used to perform pattern matching in a historical fault database, and the method of the present application includes: Using the virtual aggregate index item, a first feature pointer corresponding to the first type of potential fault mode and a second feature pointer corresponding to the second type of potential fault mode are set through association calculation; a first fault mode matching set is searched through the first feature pointer corresponding to the first type of potential fault mode; a second fault mode matching set is searched through the second feature pointer corresponding to the second type of potential fault mode; the first fault mode matching set and the second fault mode matching set are mapped to a state transition space to generate a state transition trajectory.

[0051] Specifically, the virtual aggregate index item refers to the index item dynamically generated based on the key features of the current fault mode during the pattern matching process, which is used to quickly locate records in the historical fault database that are similar to the current fault mode, thereby improving matching efficiency; correlation calculation refers to evaluating the degree of similarity between the current fault mode and the historical fault data through mathematical methods (such as similarity calculation, distance measurement, etc.). The higher the correlation, the greater the similarity between the two, which is usually based on the distance between feature vectors (such as Euclidean distance, cosine similarity, etc.); the first feature pointer / second feature pointer refers to an index pointing to a specific fault mode generated in the historical fault database based on the correlation calculation results. The first feature pointer corresponds to a type of potential fault mode, and the second feature pointer corresponds to a type of potential fault mode. It is used to quickly locate the historical record that is most similar to the current fault mode.

[0052] The fault pattern matching set refers to a set of historical records similar to the current fault pattern that are searched in the historical fault database through feature pointers. The first fault pattern matching set corresponds to one type of potential fault pattern, and the second fault pattern matching set corresponds to two types of potential fault patterns. The state transition space is a mathematical model or data structure used to describe the conversion relationship between fault modes. It records the state changes of the fault mode at different time points and reflects the development trend and evolution process of the fault. The state transition trajectory refers to the evolution path of the fault mode from the initial state to the current state in the state transition space. By analyzing the state transition trajectory, the future development trend of the fault can be predicted, providing a basis for fault diagnosis and optimization.

[0053] Using virtual aggregate index items, the similarity between the first and second potential fault modes and the historical fault data is evaluated through association calculation (such as calculating the similarity or distance between feature vectors); according to the association calculation result, a first feature pointer and a second feature pointer are set for the first and second potential fault modes respectively, pointing to the records in the historical fault database that are most similar to the current fault mode; through the first feature pointer, the historical records similar to the first potential fault mode are searched in the historical fault database to generate a first fault pattern matching set; through the second feature pointer, the historical records similar to the second potential fault mode are searched in the historical fault database to generate a second fault pattern matching set.

[0054] The first fault pattern matching set and the second fault pattern matching set are mapped to the state transfer space, which records the conversion relationship and evolution process between the fault modes; in the state transfer space, the state transfer trajectory of the fault mode is generated according to the historical records in the matching set, and the state transfer trajectory reflects the evolution path of the fault mode from the initial state to the current state, providing a basis for fault diagnosis and optimization; through correlation calculation and feature pointer setting, the system can quickly locate the historical records most similar to the current fault mode, improve the accuracy and efficiency of fault diagnosis, and the generation of state transfer trajectory provides support for dynamic analysis of faults. By analyzing the evolution path of the fault mode, the future development trend of the fault can be predicted, and preventive measures can be taken in advance to reduce the impact of the fault on the operation of campus facilities.

[0055] Furthermore, the application method includes: By calculating the correlation, the first characteristic pointer corresponding to the potential failure mode is set ;in, is the first feature pointer, D is the historical fault database, represents the feature vector corresponding to the first fault pattern matching set, is the feature vector corresponding to a class of potential failure modes, The Gaussian kernel standard deviation used to characterize the association of the first fault pattern matching set is used to control the decay speed of the similarity.

[0056] Specifically, is the first feature pointer, indicating a potential fault mode in the historical fault database D. the index or location of the most similar failure mode; Search in the historical fault database D and find the The maximum value ; Used to calculate the current potential failure mode and historical failure modes The similarity between is the feature vector corresponding to a class of potential failure modes, is a fault mode feature vector in the historical fault database D, is the Euclidean norm, which means and The distance between It refers to the standard deviation of the Gaussian kernel used to characterize the association of the first fault mode matching set. When using the Gaussian kernel for similarity calculation, it is used to control the decay rate of the similarity. The larger the standard deviation, the slower the decay rate of the similarity, that is, the gentler the decline of the similarity. Conversely, the smaller the standard deviation, the faster the decay rate of the similarity, that is, the steeper the decline of the similarity.

[0057] Use the feature vector of the current potential fault mode to compare with the feature vector in the historical fault database, and further, set the first feature pointer corresponding to the potential fault mode by calculating the correlation degree. Through the calculation of correlation and the setting of the first characteristic pointer, the historical record most similar to the current fault mode can be quickly located to improve the accuracy and efficiency of fault diagnosis. The introduction of Gaussian kernel standard deviation makes the similarity calculation more flexible and can adjust the decay speed of similarity according to actual needs, thereby more accurately reflecting the similarity between fault modes.

[0058] In summary, the beneficial effects of the embodiments of the present application are: The multimodal perception data of each park facility in the target park is collected, and a fusion data matrix is ​​generated by alignment processing, and is input into a pre-trained adaptive association model to extract the operation characteristics of the facility and construct a facility state evolution map; according to the facility health index and energy consumption anomaly coefficient, the fault risk probability and energy consumption anomaly coefficient in the facility state evolution map, the first monitoring period and the second monitoring period are configured respectively, and the first type of potential fault mode and the second type of potential fault mode are identified under the initial fault identification rule, and pattern matching is performed with the historical fault database to optimize the monitoring period and fault identification rules of each park facility with the matching results, and a fault monitoring report is output. This application provides a park facility operation status monitoring method and system based on multimodal perception, constructs a facility state evolution map, combines the time prediction of the health index and the spatial annotation of the fault risk probability, and identifies hidden faults in advance. At the same time, through multimodal data backtracking analysis and state transfer feedback optimization, the sensitivity of fault feature extraction is improved, and incremental learning and model iteration are used to continuously optimize the adaptive association model to improve the accuracy and comprehensiveness of fault identification. Technical effect.

[0059] Embodiment 2 Based on the same inventive concept as the method for monitoring the operation status of park facilities based on multimodal perception in the aforementioned embodiment, Figure 2 As shown, the embodiment of the present application provides a park facility operation status monitoring system based on multimodal perception, wherein the system includes: The data acquisition module M100 is used to collect multimodal perception data of each park facility in the target park, generate a fusion data matrix through alignment processing, and input it into the pre-trained adaptive association model to extract the operation characteristics of the facilities.

[0060] The map construction module M200 is used to construct a facility state evolution map based on the facility operation characteristics, including the facility health index, energy consumption abnormality coefficient and failure risk probability.

[0061] The first identification module M300 is used to configure a first monitoring cycle according to the facility health index and energy consumption anomaly coefficient in the facility state evolution map, and identify a type of potential fault mode under the initial fault identification rule.

[0062] The second identification module M400 is used to configure a second monitoring cycle according to the fault risk probability and energy consumption abnormality coefficient in the facility state evolution map, and identify two types of potential fault modes under the initial fault identification rules.

[0063] The pattern matching module M500 is used to perform pattern matching on the first monitoring cycle and the first type of potential fault mode, the second monitoring cycle and the second type of potential fault mode and the historical fault database, so as to optimize the monitoring cycle and fault identification rules of each park facility based on the matching results, and output a fault monitoring report.

[0064] Furthermore, the data acquisition module M100 is also used to execute the following method: The multimodal sensing data includes operating voiceprint signals, infrared thermal imaging spectra and current harmonic characteristics.

[0065] Non-intrusive current sensors are deployed in the power supply circuits corresponding to the facilities in the target park to obtain the current harmonic characteristics including the odd harmonic amplitude and phase offset.

[0066] Obtain the surface temperature field of park facilities and generate infrared thermal imaging maps with timestamps.

[0067] Use a distributed acoustic sensor array to collect operating voiceprint signals.

[0068] Furthermore, the data acquisition module M100 is also used to execute the following method: According to the operation timeline of the park facilities, align the sampling time windows of the operation voiceprint signal, infrared thermal imaging spectrum and current harmonic characteristics.

[0069] Based on the aligned sampling time windows, the phase delay errors between the multimodal data are eliminated, and the data are mapped into the feature space to generate the fused data matrix.

[0070] Furthermore, the graph construction module M200 is used to execute the following method: Spatial correlation features in the fused data matrix are extracted.

[0071] Based on the spatial correlation features in the fused data matrix, the temporal evolution trend of the facility health index is predicted.

[0072] The failure risk probability is determined through the prediction results and marked on the topological node of the facility state evolution map.

[0073] Furthermore, the first identification module M300 is also used to execute the following method: A first threshold range corresponding to a facility health index and an energy consumption anomaly coefficient is configured, and the first threshold range is adaptively adjusted according to the ambient temperature and the load rate.

[0074] In the initial fault identification rule, a first threshold range and a first monitoring period are used to establish a first state assessment decision chain corresponding to a type of potential fault mode, and the facility operation data is monitored in real time.

[0075] The first state assessment decision chain is used to perform multimodal data backtracking analysis to extract multiple abnormal feature fragments of the fault occurrence.

[0076] According to the multiple abnormal feature fragments, state transfer feedback optimization is performed on the first state assessment decision chain.

[0077] Furthermore, the pattern matching module M500 is also used to execute the following method: Configure the second threshold range corresponding to the facility health index and energy consumption abnormality coefficient.

[0078] In the initial fault identification rule, a second threshold range and a second monitoring period are used to establish a second state evaluation decision chain corresponding to the second type of potential fault mode.

[0079] The first state evaluation decision chain and the second state evaluation decision chain are used to configure a bidirectional traversal architecture, where the bidirectional traversal architecture is used to perform cross-validation with a first type of potential failure mode and a second type of potential failure mode.

[0080] Furthermore, the pattern matching module M500 is also used to execute the following method: When the first type of potential failure mode and the second type of potential failure mode are triggered simultaneously, a priority sorting mechanism of the bidirectional traversal architecture is activated.

[0081] Under the priority sorting mechanism, a virtual aggregate index item is added.

[0082] The virtual aggregate index item is used to perform pattern matching in the historical fault database.

[0083] Furthermore, the pattern matching module M500 is also used to execute the following method: The virtual aggregate index item is used to set a first characteristic pointer corresponding to the first type of potential failure mode and a second characteristic pointer corresponding to the second type of potential failure mode through association calculation.

[0084] The first fault mode matching set is searched through the first feature pointer corresponding to the first type of potential fault mode; and the second fault mode matching set is searched through the second feature pointer corresponding to the second type of potential fault mode.

[0085] The first fault mode matching set and the second fault mode matching set are mapped to a state transition space to generate a state transition trajectory.

[0086] Furthermore, the pattern matching module M500 is also used to execute the following method: By calculating the correlation, the first characteristic pointer corresponding to the potential failure mode is set .

[0087] in, is the first feature pointer, D is the historical fault database, represents the feature vector corresponding to the first fault pattern matching set, is the feature vector corresponding to a class of potential failure modes, The Gaussian kernel standard deviation used to characterize the association of the first fault pattern matching set is used to control the decay speed of the similarity.

[0088] In summary, any step can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor, and no unnecessary restrictions are made here.

[0089] Furthermore, the above-mentioned technical scheme only reflects the preferred technical scheme of the technical scheme of the embodiment of the present application. Some changes that may be made to some parts thereof by technicians in this technical field all reflect the novel principles of the embodiment of the present application. Obviously, technicians in this field can make various changes and modifications to the present application without departing from the scope of the present application.

Claims

1. A method for monitoring the operation status of park facilities based on multimodal perception, characterized in that: The method comprises: Collect multimodal perception data of each park facility in the target park, generate a fusion data matrix through alignment processing, and input it into the pre-trained adaptive association model to extract the operation characteristics of the facilities; Based on the operation characteristics of the facility, a facility status evolution map is constructed, including the facility health index, energy consumption abnormality coefficient and failure risk probability; According to the facility health index and the energy consumption abnormality coefficient in the facility state evolution map, a first monitoring cycle is configured, and a class of potential failure modes is identified under an initial failure identification rule; According to the fault risk probability and energy consumption abnormality coefficient in the facility state evolution map, a second monitoring cycle is configured, and two types of potential fault modes are identified under the initial fault identification rule; The first monitoring cycle and the first type of potential failure mode, the second monitoring cycle and the second type of potential failure mode are pattern matched with the historical failure database, and the matching results are used to optimize the monitoring cycle and fault identification rules of each park facility, and output a fault monitoring report.

2. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 1, characterized in that: The multimodal sensing data includes operating voiceprint signals, infrared thermal imaging spectra and current harmonic characteristics; Deploy non-intrusive current sensors in the power supply circuits corresponding to the facilities in the target park to obtain current harmonic characteristics including odd harmonic amplitudes and phase offsets; Obtain the surface temperature field of the park facilities and generate infrared thermal imaging maps with timestamps; Use a distributed acoustic sensor array to collect operating voiceprint signals.

3. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 2, characterized in that: Collecting multimodal sensing data of various park facilities in the target park and generating a fusion data matrix through alignment processing, the method includes: According to the operation timeline of the park facilities, align the sampling time windows of the operation voiceprint signal, infrared thermal imaging spectrum and current harmonic characteristics; Based on the aligned sampling time windows, the phase delay errors between the multimodal data are eliminated, and the data are mapped into the feature space to generate the fused data matrix.

4. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 1, characterized in that: Based on the facility operation characteristics, a facility state evolution map is constructed, the method comprising: Extracting spatial correlation features in the fused data matrix; Predicting the temporal evolution trend of the facility health index based on the spatial correlation features in the fused data matrix; The failure risk probability is determined through the prediction results and marked on the topological node of the facility state evolution map.

5. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 1, characterized in that: According to the facility health index and the energy consumption abnormality coefficient in the facility state evolution map, a first monitoring cycle is configured, and a class of potential failure modes is identified under the initial failure identification rule. The method further includes: A first threshold range corresponding to a facility health index and an energy consumption anomaly coefficient is configured, and the first threshold range is adaptively adjusted according to the ambient temperature and the load rate; In the initial fault identification rule, a first threshold range and a first monitoring period are used to establish a first state assessment decision chain corresponding to a class of potential fault modes, and the facility operation data is monitored in real time; Using the first state assessment decision chain, performing multimodal data backtracking analysis to extract multiple abnormal feature fragments of the fault; According to the multiple abnormal feature fragments, state transfer feedback optimization is performed on the first state assessment decision chain.

6. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 5, characterized in that: The first monitoring period and the first type of potential failure mode, the second monitoring period and the second type of potential failure mode are pattern matched with a historical failure database, the method comprising: Configure the second threshold range corresponding to the facility health index and energy consumption abnormality coefficient; In the initial fault identification rule, a second threshold range and a second monitoring period are used to establish a second state evaluation decision chain corresponding to the second type of potential fault mode; The first state evaluation decision chain and the second state evaluation decision chain are used to configure a bidirectional traversal architecture, where the bidirectional traversal architecture is used to perform cross-validation with a first type of potential failure mode and a second type of potential failure mode.

7. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 6, characterized in that: The bidirectional traversal architecture is used to perform cross-validation with a first type of potential failure mode and a second type of potential failure mode, and the method includes: When the first type of potential failure mode and the second type of potential failure mode are triggered simultaneously, activating a priority sorting mechanism of the bidirectional traversal architecture; Under the priority sorting mechanism, adding a virtual aggregate index item; The virtual aggregate index item is used to perform pattern matching in the historical fault database.

8. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 7, characterized in that: Using the virtual aggregate index item, pattern matching is performed in a historical fault database, and the method includes: Using the virtual aggregate index item, by calculating the degree of association, a first feature pointer corresponding to the first type of potential fault mode and a second feature pointer corresponding to the second type of potential fault mode are set; Searching for a first fault mode matching set through a first feature pointer corresponding to the first type of potential fault mode; searching for a second fault mode matching set through a second feature pointer corresponding to the second type of potential fault mode; The first fault mode matching set and the second fault mode matching set are mapped to a state transition space to generate a state transition trajectory.

9. The method for monitoring the operation status of park facilities based on multimodal perception according to claim 8, characterized in that: By calculating the correlation, the first characteristic pointer corresponding to the potential failure mode is set ; in, is the first feature pointer, D is the historical fault database, represents the feature vector corresponding to the first fault pattern matching set, is the feature vector corresponding to a class of potential failure modes, The Gaussian kernel standard deviation used to characterize the association of the first fault pattern matching set is used to control the decay speed of the similarity.

10. The park facility operation status monitoring system based on multimodal perception is characterized by: A system for implementing the park facility operation status monitoring method based on multimodal perception according to any one of claims 1 to 9, the system comprising: The data acquisition module is used to collect multimodal perception data of each park facility in the target park, generate a fusion data matrix through alignment processing, and input it into the pre-trained adaptive association model to extract the operation characteristics of the facilities; A graph construction module, for constructing a facility state evolution graph based on the facility operation characteristics, including a facility health index, an energy consumption anomaly coefficient, and a failure risk probability; A first identification module is used to configure a first monitoring period according to the facility health index and the energy consumption abnormality coefficient in the facility state evolution map, and identify a class of potential fault modes under the initial fault identification rule; A second identification module is used to configure a second monitoring period according to the fault risk probability and energy consumption abnormality coefficient in the facility state evolution map, and identify two types of potential fault modes under the initial fault identification rule; The pattern matching module is used to perform pattern matching on the first monitoring cycle and the first type of potential fault mode, the second monitoring cycle and the second type of potential fault mode and the historical fault database, so as to optimize the monitoring cycle and fault identification rules of each park facility based on the matching results, and output a fault monitoring report.

Images