Detection method of communication machine room and related equipment

Abstract

The embodiment of the invention provides a detection method of a communication machine room and related equipment, and belongs to the technical field of machine room equipment monitoring. The method comprises the following steps: acquiring machine room environment data respectively collected by various environment sensors; performing dynamic weighted fusion calculation on the sensing data of each machine room to obtain a total environment feature vector; analyzing the total environment feature vector through a lightweight model to obtain a machine room detection result; the lightweight model is obtained by learning the trained large detection model, and the lightweight model is deployed in the edge server. According to the embodiment of the invention, the lightweight model can maintain high analysis precision and generalization ability while remarkably reducing the occupation of calculation and storage resources, and finally outputs an accurate detection result of the machine room environment abnormity or the equipment state, thereby effectively improving the real-time performance and reliability of machine room environment monitoring, and improving the reliability of the machine room environment monitoring. And the autonomous decision-making and response capability at the edge side is also enhanced.

Description

Technical Field

[0001] This application relates to the field of computer room equipment monitoring technology, and in particular to a detection method and related equipment for a communication computer room. Background Technology

[0002] As core infrastructure supporting the operation of critical industries such as the internet, telecommunications, and finance, communication equipment rooms integrate servers, network equipment, storage systems, and a series of complex systems including power supply, cooling, and security. They undertake the data transmission, storage, and processing tasks of critical business operations, and have a decisive impact on business continuity and service quality. However, these facilities face multiple operational risks. The main sources of failure, in descending order of proportion, include environmental factors, power system problems, network equipment malfunctions, human error, and other factors. Once these risks occur, they can all lead to service interruptions, resulting in serious economic and social impacts. Summary of the Invention

[0003] The main objective of this application is to propose a detection method and related equipment for communication equipment rooms, which not only effectively improves the real-time performance and reliability of equipment room environment monitoring, but also enhances the autonomous decision-making and response capabilities at the edge.

[0004] To achieve the above objectives, one aspect of this application proposes a method for detecting a communication equipment room, the method comprising: Acquire data on the computer room environment collected by various environmental sensors; Dynamically weighted and fused the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector; The data center detection results are obtained by analyzing the overall environmental feature vector using a lightweight model; the lightweight model is learned from a trained large-scale detection model and is deployed on the edge server.

[0005] In some embodiments, before performing dynamic weighted fusion calculation on the sensor data of each of the computer rooms to obtain the total environmental feature vector, the method further includes: The computer room environment data collected by the various sensors are subjected to data cleaning, outlier handling, data synchronization, and data standardization to obtain the processed computer room sensor data. The step of dynamically weighting and fusing the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector includes: Dynamic weighted fusion calculations are performed on the processed data from each computer room sensor to obtain the total environmental feature vector.

[0006] In some embodiments, the step of dynamically weighting and fusing the sensor data from each of the computer rooms to obtain a total environmental feature vector includes: The sensor data of each of the aforementioned computer rooms are identified, and corresponding initial weight parameters are assigned according to the degree of influence of each computer room sensor data on the stability of the computer room. Acquire the accuracy data corresponding to each environmental sensor; Based on the accuracy data corresponding to each environmental sensor, the initial weight parameters corresponding to the data collected by each environmental sensor in the computer room are adjusted to obtain the target weight parameters corresponding to each data collected in the computer room. The total environmental feature vector is obtained by weighting the sensor data of each computer room and the target weight parameters corresponding to each sensor data of each computer room.

[0007] In some embodiments, adjusting the initial weight parameters corresponding to the data center sensor data collected by each environmental sensor based on the accuracy data corresponding to each environmental sensor, to obtain the target weight parameters corresponding to each data center sensor data, includes: If a decrease in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is reduced to obtain the target weight parameter. If an increase in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is increased to obtain the target weight parameter.

[0008] In some embodiments, the electronic device is communicatively connected to a backup power supply, a fire suppression system, and an alarm buzzer; after analyzing the overall environmental feature vector using a lightweight model to obtain the data center detection results, the method further includes: The computer room detection results are classified and detected. If the computer room detection result is a level one alarm, the backup generator is started and the fire protection system is triggered. If the detection result of the computer room is a level 2 alarm, then control the alarm buzzer to sound an intermittent alarm; If the detection result of the data center is a level 3 alarm, the data center detection result will be displayed on the target interface.

[0009] In some embodiments, the step of analyzing the total environmental feature vector using a lightweight model to obtain the data center detection results includes: The results of the neural network and the long short-term memory network in the lightweight model are obtained by reasoning through the neural network and the long short-term memory network respectively. The results of the neural network and the results of the long short-term memory network are fused to obtain the data center detection results.

[0010] In some embodiments, before acquiring the data from the computer room environment collected by various environmental sensors, the method further includes: Obtain a sample data center environment dataset; the sample data center environment dataset contains multiple sample data centers and the target data center detection results corresponding to the sample data centers. The sample data center environment dataset is input into a large-scale detection model to be trained. The large-scale detection model analyzes the sample data center environment data to obtain the predicted data center detection results corresponding to each sample data center environment. Based on the error between the predicted data center detection results corresponding to each sample data center environment and the target data center detection, the model parameters of the large-scale detection model are adjusted. If the error between the predicted data center detection result and the target data center detection corresponding to each of the sample data center environments is less than a preset error threshold, then a large-scale detection model that has been trained is obtained. The convolutional kernels of the trained large detection model with a contribution value lower than a preset contribution threshold are deleted to obtain a processed large detection model. The processed large-scale detection model is then subjected to extraction and learning to obtain a lightweight model.

[0011] To achieve the above objectives, another aspect of this application provides a detection device for a communication equipment room, the device comprising: The data acquisition module is used to acquire data on the computer room environment collected by various environmental sensors. The vector calculation module is used to perform dynamic weighted fusion calculation on the sensor data of each of the computer rooms to obtain the total environmental feature vector. The vector analysis module is used to analyze the total environmental feature vector through a lightweight model to obtain the data center detection results; the lightweight model is learned from the trained large detection model and is deployed on the edge server.

[0012] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0013] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0014] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the methods described above. The embodiments of this application include at least the following beneficial effects: This application provides a method, device, electronic device, storage medium, and program product for detecting communication equipment rooms. This solution acquires equipment room environmental data collected by various environmental sensors; performs dynamic weighted fusion calculation on the sensor data of each equipment room to obtain a total environmental feature vector; analyzes the total environmental feature vector through a lightweight model to obtain the equipment room detection result; the lightweight model is learned from a large-scale detection model that has been trained, and the lightweight model is deployed on the edge server. Implementing the embodiments of this application enables dynamic weighted fusion calculation of sensor data from various data centers, thereby generating a total environmental feature vector that comprehensively and accurately reflects the overall environmental status of the data center. This total environmental feature vector is then input into a lightweight analysis model deployed on an edge server for processing. The lightweight model is obtained by knowledge transfer and structural optimization of a fully trained large-scale detection model. While significantly reducing the consumption of computing and storage resources, it still maintains high analytical accuracy and generalization ability, ultimately outputting accurate detection results for data center environmental anomalies or equipment status. This not only effectively improves the real-time performance and reliability of data center environmental monitoring but also enhances the autonomous decision-making and response capabilities at the edge. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application; Figure 2 This is a flowchart of the detection method for a communication equipment room provided in the embodiments of this application; Figure 3 This is a schematic diagram of the hardware architecture in one embodiment; Figure 4 This is a schematic diagram of the software layer algorithm in one embodiment; Figure 5 This is a schematic diagram of the structure of the detection device for the communication equipment room provided in the embodiments of this application; Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0017] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0018] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] In related technologies, traditional manual inspections rely on maintenance personnel to conduct regular on-site inspections, visually observing equipment indicator lights, reading physical instruments, and manually recording information. This approach has inherent limitations in efficiency and real-time performance. With the application of sensor technology, single-function monitoring devices are becoming increasingly common, such as temperature and humidity sensors or smoke detectors. These devices can monitor specific parameters and trigger local alarms when preset thresholds are exceeded, but their functions are relatively isolated and data integration is lacking. Furthermore, to address the problem of data fragmentation, centralized monitoring systems have emerged. These systems combine front-end data collectors, communication networks, and back-end management platforms to achieve unified aggregation, display, and management of data from multiple types of devices.

[0021] However, these existing solutions generally face common challenges such as insufficient real-time performance, limited intelligence, inadequate data utilization, weak cross-system collaboration capabilities, incomplete security protection, and difficulty in balancing cost and benefit. Specifically, most systems cannot achieve millisecond-level fault detection and response, rely on simple threshold judgments and lack AI-based predictive capabilities, fail to fully integrate multi-dimensional data from the environment, equipment, and security to form a comprehensive health profile, lack collaborative handling mechanisms between subsystems, and exhibit disconnect between security protection and business monitoring. Furthermore, the high cost of high-performance systems and the limited functionality of low-cost solutions make it difficult to achieve both efficiency and cost-effectiveness in overall operation and maintenance. Therefore, developing a new type of intelligent operation and maintenance terminal deployed at the edge, capable of integrating multi-modal data, incorporating AI-based fault prediction, supporting cross-protocol device collaboration, and integrating security functions has become an important direction for the evolution of communication equipment room operation and maintenance models.

[0022] In view of this, this application provides a detection method for communication equipment rooms. This method not only effectively improves the real-time performance and reliability of equipment room environment monitoring, but also enhances the autonomous decision-making and response capabilities at the edge.

[0023] The communication equipment room detection method provided in this application relates to the field of equipment monitoring technology for equipment rooms. The communication equipment room detection method provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the communication equipment room detection method, but is not limited to the above forms.

[0024] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0025] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0026] like Figure 1 The diagram shown is a schematic representation of an implementation environment provided in an embodiment of this application. (Refer to...) Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected via a network, either wirelessly or via a wired connection, to complete data transmission and exchange.

[0027] Server 101 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0028] Additionally, server 101 can also be a node server in a blockchain network. Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms.

[0029] Terminal 102 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. Terminal 102 and server 101 can be directly or indirectly connected via wired or wireless communication, and this embodiment of the application does not impose any limitations.

[0030] For example, based on Figure 1 The implementation environment shown in this application embodiment provides a method for detecting a communication equipment room. The following description uses the application of this method to server 101 as an example. It can be understood that this method for detecting a communication equipment room can also be applied to terminal 102.

[0031] Figure 2 This is an optional flowchart of the detection method for a communication equipment room provided in the embodiments of this application. The subject executing the detection method for the communication equipment room can be any of the aforementioned electronic devices (including servers or terminals). Figure 2 The method may include, but is not limited to, steps S201 to S203.

[0032] Step S201: Obtain the computer room environment data collected by various environmental sensors.

[0033] In some embodiments, a variety of dedicated environmental sensors, such as precision temperature and humidity sensors, smoke detectors, airflow monitoring devices, and dust concentration sensors, can be deployed and coordinated in the communication equipment room to continuously collect multi-dimensional environmental parameters within the equipment room. These sensors, through standardized wired or wireless communication protocols, convert key physical quantities such as temperature, humidity, smoke concentration, airflow speed, and air quality monitored in real time into unified digital signals and transmit them to the data acquisition node.

[0034] By acquiring data from various environmental sensors, the data collected in the computer room forms the perception foundation of the entire intelligent operation and maintenance system. This not only overcomes the limitations of traditional manual inspections in terms of data granularity and continuity, but also provides high-quality, real-time raw data input for subsequent data fusion and in-depth intelligent analysis, thus laying a solid data foundation for achieving accurate perception and predictive maintenance of the computer room environment.

[0035] Step S202: Perform dynamic weighted fusion calculation on the sensor data of each computer room to obtain the total environmental feature vector.

[0036] In some embodiments, the weighting coefficients of each sensor parameter can be dynamically allocated and adjusted through an adaptive algorithm based on the degree of influence of each sensor parameter on the overall operating status of the computer room and the real-time quality and reliability of its data.

[0037] As an optional implementation method, before performing dynamic weighted fusion calculation on the sensor data of each computer room to obtain the total environmental feature vector, data cleaning, outlier handling, data synchronization, and data standardization processing can be performed on the computer room environmental data collected by various sensors to obtain the processed computer room sensor data; dynamic weighted fusion calculation is then performed on the processed computer room sensor data to obtain the total environmental feature vector.

[0038] In the data cleaning and preprocessing stage, the raw data collected by the sensors is first processed for missing values. If the missing data is continuous for less than 5 seconds (i.e., one sampling period), forward padding is used to fill in the missing values. If the continuous missing data exceeds 5 seconds, the system automatically switches to backup sensor data to replace the missing data, ensuring data continuity. For outlier identification, the IQR interquartile range is used. Values ​​outside the range of Q1-1.5IQR to Q3+1.5IQR are identified as outliers and replaced with the arithmetic mean of the first 10 normal data from that sensor to smooth out data fluctuations. To unify the acquisition frequency of multi-source data, high-frequency signals such as 10Hz current data are downsampled to align with 1Hz temperature and humidity data in terms of timestamps, achieving time synchronization of multi-sensor data. Finally, using the minimum and maximum values ​​of each parameter obtained from one month of historical data, all sensor parameters are Min-Max normalized, linearly mapping them to the standard range of 0 to 1 to eliminate analytical biases caused by different units of measurement, providing a high-quality, standardized data foundation for subsequent fusion calculations and intelligent analysis.

[0039] The raw environmental data collected by multiple sensors needs to undergo systematic preprocessing. This includes data cleaning using rule matching and interpolation algorithms to remove noise and invalid values; identifying and correcting abnormal values ​​caused by equipment interference or transmission errors using statistical detection or isolated forest methods; resolving timing mismatch issues caused by inconsistent acquisition frequencies of multi-source data through timestamp alignment mechanisms; and unifying various parameters to the same dimension and numerical range using normalization or Z-score standardization techniques to obtain high-quality, regularized data from the data center. Subsequently, based on the influence weight of each parameter on the system state and its data quality assessment results, an adaptive weighting strategy and feature-level fusion method are used to dynamically weight and fuse the processed data center sensor data to obtain the total environmental feature vector.

[0040] Data cleaning, outlier handling, data synchronization, and data standardization are performed on the data center environmental data collected by various sensors to obtain the processed data of each data center sensor. This effectively improves the reliability and consistency of the original data, provides high-value, integrated input features for subsequent intelligent analysis, and significantly enhances the accuracy of environmental state perception and the reliability of system decision-making.

[0041] In some embodiments, sensor data from each data center is identified, and corresponding initial weight parameters are assigned based on the degree of influence of each data center sensor data on the stability of the data center; accuracy data corresponding to each environmental sensor is obtained; based on the accuracy data corresponding to each environmental sensor, the initial weight parameters corresponding to the data center sensor data collected by each environmental sensor are adjusted to obtain the target weight parameters corresponding to each data center sensor data; and a weighted calculation is performed based on each data center sensor data and the target weight parameters corresponding to each data center sensor data to obtain the total environmental feature vector.

[0042] The system identifies and evaluates data collected from various sensors in the computer room environment, assigning corresponding initial weight parameters based on their importance in maintaining computer room stability. Simultaneously, it acquires the accuracy data of each environmental sensor, including calibration records, measurement error ranges, and other reliability indicators. Based on this accuracy data, the assigned initial weights are dynamically adjusted and optimized to obtain target weight parameters that more accurately reflect the reliability of the sensor data. Finally, based on the sensor data from each computer room and their corresponding target weight parameters, a weighted fusion algorithm is used for integrated calculation to generate a comprehensive and accurate overall environmental feature vector that fully and accurately characterizes the overall environmental state of the computer room.

[0043] Based on the criticality of each environmental parameter to the stability of the data center operation, a basic weight is assigned. This weight is then dynamically adjusted in conjunction with real-time sensor accuracy data. The final weight used for fusion calculation is obtained by multiplying the basic weight by an accuracy coefficient reflecting the current measurement reliability. During the fusion analysis phase, the standardized values ​​of each parameter are weighted and summed with their final weights to calculate an overall fusion score ranging from 0 to 1. Risk levels are then categorized based on this score: a score above 0.7 is classified as high-risk (P1), typically corresponding to multiple serious anomalies; a score between 0.4 and 0.7 is classified as medium-risk (P2), indicating a combination of anomalies requiring attention; and a score below 0.4 is classified as low-risk (P3), reflecting localized or minor anomalies. This mechanism, through quantitative assessment and grading, provides data center operation and maintenance with precise status awareness and tiered early warning capabilities.

[0044] By integrating multi-source sensor information and dynamically optimizing weight allocation, the accuracy and reliability of environmental status assessment are effectively improved, providing a high-quality data foundation for subsequent intelligent analysis and decision-making.

[0045] Furthermore, if a decrease in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is reduced to obtain the target weight parameter; if an increase in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is increased to obtain the target weight parameter. The system can continuously monitor the accuracy data of each environmental sensor and adjust its weight in the fusion process in real time: when a decrease in the accuracy data of a certain environmental sensor is detected, indicating a decrease in its measurement reliability, the system will reduce the initial weight parameter corresponding to the data collected by that sensor through a weight adaptive adjustment algorithm. Conversely, if the sensor's accuracy data increases, its weight is increased according to the same mechanism.

[0046] Dynamically adjusting the weight parameters based on sensor accuracy feedback can effectively improve the accuracy and robustness of multi-source data fusion, ensuring that the final generated total environmental feature vector more accurately reflects the actual operating status of the data center.

[0047] Step S203: Analyze the total environmental feature vector using a lightweight model to obtain the data center detection results; the lightweight model is learned from the trained large-scale detection model and is deployed on the edge server.

[0048] In some embodiments, the data is input into a lightweight analysis model deployed on an edge server for deep analysis to obtain comprehensive data center environment detection results. This lightweight model is an embedded version obtained by employing model compression techniques such as knowledge distillation, through knowledge transfer and parameter simplification of a large, high-precision detection model pre-trained on a high-performance server. It inherits the core detection capabilities and judgment logic of the original model, but significantly reduces computational complexity and resource consumption.

[0049] Furthermore, inference is performed separately using the neural network and long short-term memory network in the lightweight model to obtain neural network results and long short-term memory network results, respectively. These results are then fused to obtain the data center detection result. The preprocessed and fused total environmental feature vector is input in parallel into the two core network structures within the lightweight model deployed on the edge server for collaborative inference. The feedforward neural network extracts deep features from the current input and generates neural network results based on the instantaneous state, while the long short-term memory network, relying on its unique gating mechanism and state memory units, models the temporal dependencies in the feature vector and outputs long short-term memory network results reflecting historical trends. Subsequently, model fusion techniques such as weighted averaging or feature concatenation are used to effectively integrate the two inference results representing the current state and time trends, respectively, ultimately generating a comprehensive data center detection result.

[0050] By performing inference using the neural network and long short-term memory network in the lightweight model, the results of the neural network and the long short-term memory network are obtained. The results of the neural network and the long short-term memory network are then fused to obtain the data center detection results. This not only comprehensively captures the complex state characteristics of the data center environment, but also significantly improves the accuracy of fault early warning and the system's ability to perceive potential risks.

[0051] In some embodiments, the electronic device is communicatively connected to a backup power supply, a fire protection system, and an alarm buzzer. After analyzing the overall environmental feature vector using a lightweight model to obtain the data center detection results, the data center detection results can be classified. If a Level 1 alarm is detected, the backup generator is started and the fire protection system is triggered. If a Level 2 alarm is detected, the alarm buzzer is controlled to sound intermittently. If a Level 3 alarm is detected, the data center detection results are displayed on the target interface.

[0052] By using a lightweight analysis model deployed on an edge server to analyze the overall environmental feature vector in real time, and generating the data center environment status detection results, the system can perform graded responses based on preset rules: when a level 1 alarm is detected, the edge computing device will send a start command to the backup power system through the communication interface, and at the same time trigger the fire protection system to enter emergency working state; when a level 2 alarm is detected, the system will control the alarm buzzer to execute a preset intermittent sounding mode; if the detection result is a level 3 alarm, the status prompt and result display will only be displayed in a designated area of ​​the monitoring interface.

[0053] Specifically, it can execute differentiated response strategies based on the severity of the incident. When a Level 1 alarm is triggered, it will automatically cut off the power supply to non-core equipment, immediately start the backup generator, and activate the fire protection system to enter emergency response mode. For a Level 2 alarm, it will simultaneously activate the buzzer for intermittent sounding and the yellow flashing light for warning, push alarm information to the electronic devices of the operation and maintenance manager, and automatically start the ventilation system and reduce the load on non-core equipment to control risks. If it is a Level 3 alarm, it will provide on-site notification by keeping the yellow LED indicator constantly lit, push reminder information to the central management terminal, record the operation log, and automatically generate a planned maintenance work order to provide a basis for subsequent operation and maintenance decisions.

[0054] The precise tiered response strategy ensures rapid emergency response to major risks while avoiding over-processing of minor alarms, effectively improving the automation level and security capabilities of data center operation and maintenance.

[0055] In some embodiments, when an intrusion is detected by an infrared detector or an unauthorized card reading event is detected by an access control controller, a linkage mechanism is automatically triggered to control a pre-set camera to continuously capture three images at 1-second intervals. Subsequently, the captured images are subjected to real-time AI recognition based on computer vision technology, including determining whether the person is a registered maintenance personnel through facial comparison algorithms and detecting whether they are carrying suspicious items using a behavioral analysis model. Finally, a tiered response is executed based on the recognition results: if the person is identified as a non-maintenance personnel, their facial image and alarm information can be automatically pushed to the monitoring center simultaneously; if they are also identified as carrying suspicious items, the alarm level is immediately upgraded, for example, from P2 medium risk to P1 high risk, thereby realizing intelligent security closed-loop management from perception, recognition to handling.

[0056] In some embodiments, when an alarm is triggered, the terminal will record the alarm generation time, specific type, triggering conditions, and related on-site images, and immediately push the alarm details remotely to the mobile terminal of the maintenance personnel for confirmation. If a P1 level alarm is not manually confirmed within 30 minutes, it will automatically dial multiple preset maintenance personnel numbers in sequence for voice notification. After the on-site handling is completed, the maintenance personnel can submit feedback information containing specific handling measures through a dedicated APP. The terminal will then stop the alarm and generate a complete handling log, forming a full-process traceability mechanism from alarm triggering to closed-loop management.

[0057] In some embodiments, the electronic device defines three key log types for comprehensively recording operational status. The device status log continuously records various system parameters every minute; the fault log is automatically generated each time an alarm is triggered; and the operation log faithfully records all local and remote manual interventions. For storage management, a dual strategy of local and cloud collaboration is adopted. The device uses a 32GB TF card to cyclically store the complete log data for the most recent month, while key log information is synchronized in real-time to a remote operation and maintenance platform for long-term archiving for one year. To facilitate on-site auditing and analysis, structured Excel format log files can be quickly exported via USB interface. These files fully include key fields such as timestamps, device identifiers, event categories, detailed descriptions, and handling results, achieving full traceability of the operation and maintenance process.

[0058] As an optional implementation, a sample data center environment dataset can be obtained. This dataset contains multiple sample data center environments and the corresponding target data center detection results. The sample data center environment dataset is input into a large-scale detection model to be trained. The large-scale detection model analyzes each sample data center to obtain the predicted data center detection results. Based on the error between the predicted data center detection results and the target data center detection, the model parameters of the large-scale detection model are adjusted. If the error between the predicted data center detection results and the target data center detection is less than a preset error threshold, the trained large-scale detection model is obtained. Convolutional kernels of the trained large-scale detection model with contributions lower than a preset contribution threshold are deleted to obtain a processed large-scale detection model. The processed large-scale detection model is then subjected to extraction learning to obtain a lightweight model.

[0059] In the model construction and optimization process, a dataset containing 100,000 labeled samples was first constructed based on the annual operation and maintenance logs of a medium-sized communication equipment room. This dataset covers five typical fault types and is divided into training, validation, and test sets in a 7:2:1 ratio. Each sample is labeled with whether a fault occurred and its specific multi-label fault category. During the feature engineering phase, a sliding window with a length of 10 minutes and a step size of 1 minute was used to segment the original time-series data. For each window, 12 statistical features, including mean, variance, peak-to-peak value, zero-crossing rate, trend slope, and correlation, were extracted to form a 128-dimensional feature vector. Data augmentation methods such as time series flipping, adding Gaussian noise, and interpolation were used to increase the total number of samples to 200,000. For model training, the AdamW optimizer was used with weight decay to suppress overfitting. The initial learning rate was 1e-3, dynamically adjusted using a cosine annealing strategy. The batch size was set to 32 to adapt to the memory limitations of edge devices. The training consisted of 50 epochs with an early stopping mechanism. The loss function used was BCEWithLogitsLoss, suitable for multi-label classification. To meet the needs of edge deployment, the trained model was sequentially subjected to INT8 quantization to compress the model size from 25MB to 6MB while keeping the accuracy loss within 3%. Then, channel pruning guided by L1 regularization was used to further compress the model to 4.5MB. Finally, knowledge distillation technology with ResNet-LSTM as the teacher model was used to restore the model performance, so that the lightweight model achieved an F1 score of 0.92 on the test set, realizing efficient inference under limited resources.

[0060] To adapt to the limited computing resources of edge terminals, a systematic optimization is carried out on the basis of a hybrid model of neural networks and long short-term memory networks, focusing on four dimensions: model structure, parameter compression, inference acceleration, and dynamic updates, to achieve a balance between efficient operation and high-precision fault prediction at the edge.

[0061] At the structural level, a combination of a lightweight GRU (gated recurrent unit) and an attention mechanism replaces the original LSTM (long short-term memory) module. This reduces the number of parameters by 42% while retaining the ability to model long-range dependencies. Depthwise separable convolutions are introduced to replace standard convolutions, reducing the kernel parameters to one-ninth of the original and achieving a feature extraction latency of less than 30 milliseconds. For parameter compression, a joint optimization strategy of quantization, pruning, and distillation is employed. Through layer-by-layer quantization-aware training, the model is compressed from 25MB to 6MB with an accuracy loss of less than 1.5%. Structured pruning controls the pruning rate to within 30% to adapt to the ARM Neon instruction set. Furthermore, knowledge distillation technology using ResNet-LSTM as the teacher model achieves an 82% model compression rate while restoring the F1-score to 0.92.

[0062] For inference acceleration, a customized TFLite Micro runtime was implemented for the ARM Cortex-A7 architecture, enabling the XNNPACK acceleration library and fusion of INT8 operators, resulting in a single inference latency of less than 180 milliseconds. Simultaneously, a model sharding loading mechanism kept peak memory usage below 512MB. For the update mechanism, a sliding window-based incremental learning strategy was proposed. The terminal can periodically collect new data and fine-tune less than 5% of parameters locally. A forgetting detection mechanism was introduced to automatically trigger model updates when data distribution drift exceeds a threshold, achieving a long-term prediction accuracy decay of less than 1% per month. After system-wide optimization, the fault prediction model runs stably on the Allwinner V3S processor at the edge terminal with a size of 4.5MB, autonomously completing incremental learning every seven days. This represents the first self-evolving, high-precision edge intelligent fault prediction system in a communication equipment room scenario, effectively resolving the core contradiction between accuracy, speed, and resources in edge AI deployment.

[0063] Figure 3 A schematic diagram of the hardware architecture in one embodiment, such as Figure 3 As shown, the hardware layer focuses on "industrial-grade reliability + multi-module compatibility + edge computing support" as its core objectives, and covers five core units: sensor access module, device control module, main control computing module, communication transmission module, and interactive display module. The selection and design of each module are customized for the data center scenario.

[0064] The sensor access module is the "sensory nerve of the terminal," covering three core data categories: data from the server room environment, equipment, and security. All sensors are industrial-grade models, adapted to the server room's wide temperature range of -10~50℃ and strong electromagnetic interference. Industrial-grade temperature, humidity, gas, smoke, current, and voltage sensors are selected, and the camera uses a Hikvision DS-IPC-T12H high-definition camera. Environmental sensors are installed in layers according to the rack height, with equipment sensors close to the device being measured to avoid data transmission loss. Critical sensors (such as temperature, humidity, and current) employ a "dual-path" system; if the primary sensor fails, the backup sensor automatically switches over with a switching time of <100ms. All sensor cables are shielded (with the shield grounded), and analog signals (such as voltage and current) are filtered by an RC filter circuit to reduce electromagnetic interference.

[0065] The equipment control module is the "execution center of the terminal" and needs to be compatible with high-voltage, multi-protocol equipment in the computer room (such as precision air conditioners, UPS, and fans). The core is to achieve unified control through "industrial-grade execution units + protocol conversion plug-ins". The execution units adopt a relay solution, and the protocol conversion plug-ins solve the problem of fragmented protocols of computer room equipment (such as precision air conditioners using Modbus-RTU, UPS using SNMP, and switches using CLI). The protocol conversion module realizes "unified command issuance".

[0066] The communication transmission module is the "nerve center" of the terminal, and it needs to cover three communication scenarios: local area network (LAN), remote, and short-range, to ensure the reliability of data acquisition, command issuance, and alarm push. Ethernet is the primary communication link, WiFi / Zigbee is the LAN backup, and 4G is the remote backup. If the primary link is interrupted, it will automatically switch to the backup link (switching time <1s). Remote communication (4G) adopts a "data compression + timed upload" strategy (e.g., uploading historical data once every 5 minutes, with a single data size <10KB) to reduce traffic costs. All communication links use encrypted transmission (Ethernet: TLS1.3; WiFi: WPA3; 4G: IPSec) to prevent data theft or tampering.

[0067] The interactive display module serves as the "human-computer interaction window" for the terminal, requiring it to meet the needs of local operation, status viewing, and emergency control within the data center. Its design balances practicality and interference resistance. Critical operations (such as emergency shutdown and parameter modification) require "double-click confirmation" or "password verification" to prevent accidental operation. The touchscreen supports automatic brightness adjustment (50~300 cd / ㎡ based on the data center's lighting intensity) to prevent poor visibility in strong light. Even without network access, historical data (last 7 days) can still be viewed locally, and emergency control can be executed, ensuring uninterrupted operation and maintenance during network outages.

[0068] Figure 4 This is a schematic diagram of a software layer algorithm in one embodiment, such as... Figure 4 As shown, its core is the software layer - AI and algorithms. Extending outward from this core are four modules: a multimodal data fusion module for integrating multiple types of data, an AI fault prediction module responsible for predicting equipment faults in advance, a device protocol adaptation module that can adapt to different device protocols, and a security linkage and alarm module that can realize security linkage and alarm functions.

[0069] Steps S201 to S203 of this embodiment involve acquiring data from various environmental sensors in the computer room; dynamically weighting and fusing the sensor data from each computer room to obtain a total environmental feature vector; and analyzing the total environmental feature vector using a lightweight model to obtain the computer room detection result. The lightweight model is learned from a trained large-scale detection model and deployed on the edge server. It can dynamically weight and fuse the sensor data from each computer room to generate a total environmental feature vector that comprehensively and accurately reflects the overall environmental status of the computer room. This total environmental feature vector is then input into the lightweight analysis model deployed on the edge server for processing. The lightweight model is obtained through knowledge transfer and structural optimization of the trained large-scale detection model. While significantly reducing computational and storage resource consumption, it maintains high analytical accuracy and generalization ability, ultimately outputting accurate detection results for computer room environmental anomalies or equipment status. This not only effectively improves the real-time performance and reliability of computer room environmental monitoring but also enhances the autonomous decision-making and response capabilities at the edge.

[0070] Please see Figure 5 This application also provides a detection device for a communication equipment room, which can implement the above-described method. The device includes: The data acquisition module 501 is used to acquire data on the computer room environment collected by various environmental sensors. The vector calculation module 502 is used to perform dynamic weighted fusion calculation on the sensor data of each computer room to obtain the total environmental feature vector. The vector analysis module 503 is used to analyze the total environmental feature vector through a lightweight model to obtain the data center detection results. The lightweight model is learned from the trained large detection model and is deployed on the edge server.

[0071] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0072] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0073] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0074] Please see Figure 6 , Figure 6 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 601 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 602 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 602 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 602 and is called and executed by the processor 601 using the methods described in the embodiments of this application. The input / output interface 603 is used to implement information input and output; The communication interface 604 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 605 transmits information between various components of the device (e.g., processor 601, memory 602, input / output interface 603, and communication interface 604); The processor 601, memory 602, input / output interface 603, and communication interface 604 are connected to each other within the device via bus 605.

[0075] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0076] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0077] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0078] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0079] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0080] The communication equipment room detection method, device, electronic equipment, storage medium, and program product provided in this application embodiment acquire equipment room environmental data collected by various environmental sensors; dynamically weight and fuse the sensor data of each equipment room to obtain a total environmental feature vector; and analyze the total environmental feature vector through a lightweight model to obtain the equipment room detection result. The lightweight model is learned from a large-scale detection model that has been trained. The lightweight model is deployed on the edge server, which not only effectively improves the real-time performance and reliability of equipment room environmental monitoring, but also enhances the autonomous decision-making and response capabilities at the edge.

[0081] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0082] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0084] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0085] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0086] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0087] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0088] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0089] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0090] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0091] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for detecting a communication equipment room, characterized in that, Applied to an electronic device that is communicatively connected to an edge server, the method includes the following steps: Acquire data on the computer room environment collected by various environmental sensors; Dynamically weighted and fused the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector; The data center detection results are obtained by analyzing the overall environmental feature vector using a lightweight model; the lightweight model is learned from a trained large-scale detection model and is deployed on the edge server.

2. The method according to claim 1, characterized in that, Before performing dynamic weighted fusion calculation on the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector, the method further includes: The computer room environment data collected by the various sensors are subjected to data cleaning, outlier handling, data synchronization, and data standardization to obtain the processed computer room sensor data. The step of dynamically weighting and fusing the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector includes: Dynamic weighted fusion calculations are performed on the processed data from each computer room sensor to obtain the total environmental feature vector.

3. The method according to claim 1, characterized in that, The step of dynamically weighting and fusing the sensor data from each of the aforementioned computer rooms to obtain the total environmental feature vector includes: The sensor data of each of the aforementioned computer rooms are identified, and corresponding initial weight parameters are assigned according to the degree of influence of each computer room sensor data on the stability of the computer room. Acquire the accuracy data corresponding to each environmental sensor; Based on the accuracy data corresponding to each environmental sensor, the initial weight parameters corresponding to the data collected by each environmental sensor in the computer room are adjusted to obtain the target weight parameters corresponding to each data collected in the computer room. The total environmental feature vector is obtained by weighting the sensor data of each computer room and the target weight parameters corresponding to each sensor data of each computer room.

4. The method according to claim 3, characterized in that, The step of adjusting the initial weight parameters corresponding to the data collected by each environmental sensor based on the accuracy data of each environmental sensor to obtain the target weight parameters corresponding to each data of the data of the data center includes: If a decrease in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is reduced to obtain the target weight parameter. If an increase in the accuracy data corresponding to the first environmental sensor is detected, the initial weight parameter corresponding to the data collected by the first environmental sensor in the computer room is increased to obtain the target weight parameter.

5. The method according to claim 1, characterized in that, The electronic device is communicatively connected to the backup power supply, fire protection system, and alarm buzzer; after analyzing the overall environmental feature vector using a lightweight model to obtain the data center detection results, the method further includes: The computer room detection results are classified and detected. If the computer room detection result is a level one alarm, the backup generator is started and the fire protection system is triggered. If the detection result of the computer room is a level 2 alarm, then control the alarm buzzer to sound an intermittent alarm; If the detection result of the data center is a level 3 alarm, the data center detection result will be displayed on the target interface.

6. The method according to claim 1, characterized in that, The analysis of the overall environmental feature vector using a lightweight model to obtain the data center detection results includes: The results of the neural network and the long short-term memory network in the lightweight model are obtained by reasoning through the neural network and the long short-term memory network respectively. The results of the neural network and the results of the long short-term memory network are fused to obtain the data center detection results.

7. The method according to any one of claims 1 to 6, characterized in that, Before acquiring the data from the various environmental sensors that collect the computer room environment data, the method further includes: Obtain a sample data center environment dataset; the sample data center environment dataset contains multiple sample data centers and the target data center detection results corresponding to the sample data centers. The sample data center environment dataset is input into a large-scale detection model to be trained. The large-scale detection model analyzes the sample data center environment data to obtain the predicted data center detection results corresponding to each sample data center environment. Based on the error between the predicted data center detection results corresponding to each sample data center environment and the target data center detection, the model parameters of the large-scale detection model are adjusted. If the error between the predicted data center detection result and the target data center detection corresponding to each of the sample data center environments is less than a preset error threshold, then a large-scale detection model that has been trained is obtained. The convolutional kernels of the trained large detection model with a contribution value lower than a preset contribution threshold are deleted to obtain a processed large detection model. The processed large-scale detection model is then subjected to extraction and learning to obtain a lightweight model.

8. A detection device for a communication equipment room, characterized in that, The device includes: The data acquisition module is used to acquire data on the computer room environment collected by various environmental sensors. The vector calculation module is used to perform dynamic weighted fusion calculation on the sensor data of each of the computer rooms to obtain the total environmental feature vector. The vector analysis module is used to analyze the total environmental feature vector through a lightweight model to obtain the data center detection results; the lightweight model is learned from the trained large detection model and is deployed on the edge server.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images