Intelligent distribution method and system for sequential inspection of machine rooms

Abstract

The invention provides an intelligent distribution method and system for sequential inspection of a machine room, and the method comprises the steps: deploying an Internet of Things sensor and monitoring equipment in the machine room, so as to collect the operation state data of the equipment in the machine room in real time; marking the running state data by adopting the timestamp and the equipment ID of the corresponding equipment to obtain marked data; analyzing the marked data by using a machine learning algorithm so as to predict the health state of the equipment, obtaining a fault risk prediction result, and allocating a health score to each piece of equipment according to the fault risk prediction result; generating a to-be-inspected equipment list according to a preset inspection period and the health score of the equipment; based on a path optimization algorithm, generating an optimal inspection path according to the geographic position of each device in the device list; and sending the optimal inspection path to the inspection personnel to realize the intelligent distribution of the sequential inspection of the machine rooms. According to the invention, the quality and efficiency of machine room equipment management can be obviously improved, and the operation and maintenance cost and risk are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of machine room inspection, and in particular to an intelligent allocation method and system for sequential inspection of a machine room. Background Art

[0002] The hospital computer room is the core infrastructure of the medical information system, and its operating status directly affects the daily operation of the hospital. The equipment in the computer room (such as servers, storage devices, network equipment, UPS power supplies, air conditioners, etc.) needs to be inspected regularly to ensure its normal operation. However, traditional inspection methods usually rely on manual experience and have the following problems:

[0003] Low inspection efficiency: Manually assigning inspection tasks is prone to omissions or duplications, resulting in low inspection efficiency.

[0004] Unbalanced task distribution: Improper distribution of inspection tasks may result in some personnel having too much workload and others having too little workload.

[0005] Lack of real-time adjustment capabilities: When the status of the equipment in the computer room changes, the inspection plan cannot be adjusted dynamically.

[0006] Non-standard data recording: The inspection data recording method is not unified, which makes subsequent analysis and optimization difficult.

[0007] Therefore, an intelligent inspection task allocation method is proposed, which uses artificial intelligence and Internet of Things technologies to achieve efficient allocation and dynamic adjustment of inspection tasks. Summary of the invention

[0008] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an intelligent allocation method and system for sequential inspection of computer rooms. Through the combination of the Internet of Things, machine learning and path optimization technology, the inspection work is made intelligent and efficient, which can significantly improve the quality and efficiency of computer room equipment management, while reducing operation and maintenance costs and risks.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] An intelligent allocation method for sequential inspection of a computer room, comprising:

[0011] Deploy IoT sensors and monitoring equipment in the computer room to collect real-time operating status data of the equipment inside the computer room;

[0012] Marking the operation status data using a timestamp and a device ID of a corresponding device to obtain marked data;

[0013] Analyzing the labeled data using a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assigning a health score to each device according to the fault risk prediction result;

[0014] Generate a list of devices to be inspected based on the preset inspection cycle and health score of the devices;

[0015] Based on the path optimization algorithm, an optimal inspection path is generated according to the geographical location of each device in the device list;

[0016] The optimal inspection path is sent to the inspection personnel to realize the intelligent allocation of sequential inspection of the computer room.

[0017] Preferably, the operating status data includes: temperature signal, humidity signal, voltage signal, current signal, equipment operation log and alarm signal.

[0018] Preferably, the operating status data is marked using a timestamp and a device ID of a corresponding device to obtain marked data, including:

[0019] Add a timestamp and device ID to each collected operation status data record; the timestamp is used to record the specific time point of data collection, and the device ID is used to identify the specific device from which the data comes;

[0020] Integrate multiple operating status data of the same device according to timestamps to generate unified tag data; the format of the tag data is: D(t,ID)={T(t,ID),H(t,ID),V(t,ID),I(t,ID),L(t,ID),A(t,ID)}; wherein D(t,ID) represents a set of tag data collected by the device ID at time t; T(t,ID) represents the temperature signal value of the device ID at time t; H(t,ID) represents the humidity signal value of the device ID at time t; V(t,ID) represents the voltage signal value of the device ID at time t; I(t,ID) represents the current signal value of the device ID at time t; L(t,ID) represents the operating log of the device ID at time t; A(t,ID) represents the alarm signal of the device ID at time t; t represents the specific time point of data collection; ID represents the unique identifier of the device.

[0021] Preferably, after the operating status data is marked by using the timestamp and the device ID of the corresponding device to obtain the marked data, the method further includes:

[0022] The marked data are grouped according to a preset collection period to obtain a plurality of data groups;

[0023] Calculate the difference coefficient between the current data group and the previous data group in sequence;

[0024] Determine whether the value of the coefficient of difference is within a preset range;

[0025] If the value of the difference coefficient is not within the preset range, the corresponding data group is removed;

[0026] If the value of the difference coefficient is within the preset range, the corresponding data group will be retained until all data groups are traversed to obtain the labeled data after data cleaning.

[0027] Preferably, the coefficient of difference calculation formula is:

[0028]

[0029] Among them, p X,Y is the coefficient of difference, cov(X,Y) represents the covariance between the current data set X and the previous data set Y, α X represents the mean of the current data set X, β Y Represents the mean of the previous data set Y.

[0030] Preferably, the labeled data is analyzed using a machine learning algorithm to predict the health status of the equipment and obtain a fault risk prediction result, including:

[0031] Obtain a preset sample labeled data set;

[0032] Build an initial deep learning model;

[0033] Training the initial deep learning model according to the sample labeled data set to obtain a trained classifier;

[0034] The trained LSTM neural network is connected after the classifier to obtain the fault risk prediction result.

[0035] Preferably, the calculation formula of the health score is:

[0036] HS=(1-P_fault)×100-S_abnormal×W_device;

[0037] Among them, HS is the health score, P_fault is the fault risk prediction result, S_abnormal is the abnormality of the current operating state, and W_device is the importance weight of the device; the calculation formula of the abnormality of the current operating state is: S_abnormal=Σ(|X_i-X_normal| / R_i)×W_i; Xi is the operating state data currently collected by the device, X_normal is the normal value of the operating state data corresponding to the device, R_i is the normal range of the operating state data, and W_i is the weight of the operating state data.

[0038] Preferably, a list of devices to be inspected is generated based on the preset inspection data and health scores of the devices, including:

[0039] Obtain the preset inspection data; the preset inspection data includes: device ID, device name, device type, device location, inspection cycle and last inspection time;

[0040] The time interval from the last inspection is calculated according to the formula T_interval=T_now-T_last; wherein T_interval is the time interval; T_now is the current time; and T_last is the last inspection time;

[0041] The inspection urgency is calculated according to the formula U_inspect=T_interval / C_device; wherein U_inspect is the inspection urgency; C_device is the inspection period;

[0042] The inspection priority of the equipment is determined according to the formula P_inspect = W_health × (1-HS / 100) + W_time × U_inspect; W_health is the weight of the health score, W_time is the weight of the inspection urgency, and P_inspect is the inspection priority;

[0043] Determine the inspection priority threshold P_threshold according to the preset inspection resources and time limit, and add the devices with inspection priority P_inspect ≥ P_threshold to the list of devices to be inspected;

[0044] Sort the devices from high to low according to the inspection priority P_inspect, give priority to high-priority devices for inspection, and generate a final sorted list of devices to be inspected.

[0045] Preferably, the path optimization algorithm is Dijkstra algorithm.

[0046] An intelligent distribution system for sequential inspection of a computer room, comprising:

[0047] A data acquisition unit, used to deploy IoT sensors and monitoring equipment in the computer room to collect operating status data of the equipment inside the computer room in real time;

[0048] A data marking unit, used to mark the operation status data with a timestamp and a device ID of a corresponding device to obtain marked data;

[0049] A scoring calculation unit, configured to analyze the labeled data using a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assign a health score to each equipment according to the fault risk prediction result;

[0050] A list generation unit, used to generate a list of devices to be inspected according to a preset inspection cycle and health score of the devices;

[0051] A path optimization unit, configured to generate an optimal inspection path according to the geographical location of each device in the device list based on a path optimization algorithm;

[0052] The sequential inspection unit is used to send the optimal inspection path to the inspection personnel to realize the intelligent allocation of the sequential inspection of the computer room.

[0053] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0054] The present invention provides an intelligent allocation method and system for sequential inspection of a computer room, including: deploying Internet of Things sensors and monitoring equipment in the computer room to collect the operating status data of the equipment inside the computer room in real time; marking the operating status data with a timestamp and the device ID of the corresponding equipment to obtain marked data; analyzing the marked data with a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assign a health score to each device according to the fault risk prediction result; generating a list of equipment to be inspected according to the preset inspection cycle and health score of the equipment; generating an optimal inspection path based on the geographical location of each device in the equipment list based on a path optimization algorithm; sending the optimal inspection path to the inspection personnel to realize the intelligent allocation of sequential inspection of the computer room. The present invention realizes the intelligence and efficiency of inspection work through the combination of Internet of Things, machine learning and path optimization technology, which can significantly improve the quality and efficiency of equipment management in the computer room, while reducing operation and maintenance costs and risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of the system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The purpose of the present invention is to provide an intelligent allocation method and system for sequential inspection of computer rooms. Through the combination of the Internet of Things, machine learning and path optimization technology, the inspection work is made intelligent and efficient, which can significantly improve the quality and efficiency of computer room equipment management, while reducing operation and maintenance costs and risks.

[0060] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides an intelligent allocation method for sequential inspection of a computer room, comprising:

[0062] Step 100: deploying IoT sensors and monitoring equipment in the computer room to collect operating status data of the equipment in the computer room in real time;

[0063] Step 200: Mark the operation status data using a timestamp and a device ID of a corresponding device to obtain marked data;

[0064] Step 300: Analyze the labeled data using a machine learning algorithm to predict the health status of the device, obtain a fault risk prediction result, and assign a health score to each device according to the fault risk prediction result;

[0065] Step 400: Generate a list of devices to be inspected according to the preset inspection cycle and health score of the devices;

[0066] Step 500: Based on a path optimization algorithm, an optimal inspection path is generated according to the geographical location of each device in the device list;

[0067] Step 600: Send the optimal inspection path to the inspection personnel to realize the intelligent allocation of sequential inspection of the computer room.

[0068] Preferably, the operating status data includes: temperature signal, humidity signal, voltage signal, current signal, equipment operation log and alarm signal.

[0069] Specifically, step 100 of this embodiment includes:

[0070] Step 101: Deployment of sensors and monitoring equipment

[0071] In the computer room, this embodiment reasonably deploys IoT sensors and monitoring equipment according to the distribution of equipment and the structure of the environment. Sensor types include temperature and humidity sensors, current sensors, voltage sensors, etc., which are used to collect physical parameters of the equipment in real time during operation; monitoring equipment includes cameras and alarm devices to record equipment operation logs and alarm information. During the deployment process, it is necessary to ensure that the installation location of the sensor can cover all key equipment and avoid signal interference or obstruction. For example, temperature and humidity sensors should be placed in representative locations in the computer room, voltage and current sensors should be directly connected to the power lines of the equipment, and monitoring cameras should cover the main areas of the computer room to monitor the operating status of the equipment.

[0072] Step 102: Sensor and monitoring equipment connection and data collection

[0073] After deployment, this embodiment connects all sensors and monitoring devices to the central data acquisition system via wired (such as Ethernet) or wireless (such as Wi-Fi, ZigBee). Each sensor and device needs to be assigned a unique device ID for subsequent data tagging and analysis. The data collected by the sensor in real time includes physical parameters such as temperature, humidity, voltage, and current. The logs and alarm information generated by the monitoring equipment are sent to the acquisition system through the data interface. The acquisition system uses a timestamp to record the time point of each data collection to ensure the timing and accuracy of the data. In order to improve the reliability of data collection, the sampling frequency of data collection can be set (such as once a minute) and the data can be preliminarily verified to filter out abnormal values ​​or noise.

[0074] Step 103: Data storage and transmission

[0075] The collected operating status data is transmitted to the central server or cloud platform through the network for storage and processing. The storage system uses a structured database (such as MySQL) or a time series database (such as InfluxDB) to classify and store the data. The data table contains fields such as device ID, timestamp, operating status parameters (such as temperature, humidity, voltage, etc.), and alarm information. At the same time, the system must support real-time data transmission and historical data storage for subsequent fault prediction and inspection task allocation. During the data storage process, encrypted transmission and permission management mechanisms are used to ensure the security and integrity of the data. In addition, in order to facilitate subsequent analysis, a data interface (such as API) can be provided for machine learning models or inspection systems to call.

[0076] Preferably, the operating status data is marked using a timestamp and a device ID of a corresponding device to obtain marked data, including:

[0077] Add a timestamp and device ID to each collected operation status data record; the timestamp is used to record the specific time point of data collection, and the device ID is used to identify the specific device from which the data comes;

[0078] Integrate multiple operating status data of the same device according to timestamps to generate unified tag data; the format of the tag data is: D(t,ID)={T(t,ID),H(t,ID),V(t,ID),I(t,ID),L(t,ID),A(t,ID)}; wherein D(t,ID) represents a set of tag data collected by the device ID at time t; T(t,ID) represents the temperature signal value of the device ID at time t; H(t,ID) represents the humidity signal value of the device ID at time t; V(t,ID) represents the voltage signal value of the device ID at time t; I(t,ID) represents the current signal value of the device ID at time t; L(t,ID) represents the operating log of the device ID at time t; A(t,ID) represents the alarm signal of the device ID at time t; t represents the specific time point of data collection; ID represents the unique identifier of the device.

[0079] Specifically, step 200 of this embodiment includes:

[0080] Step 201: Add a timestamp and device ID to each piece of collected operating status data

[0081] During the data collection phase, IoT sensors and monitoring devices collect operating status data (such as temperature, humidity, voltage, etc.) in real time. Each sensor and device is configured with a unique device ID (ID) to identify the data source. When the system receives data, it automatically adds a timestamp (t) to each data record. The timestamp records the specific time point of data collection in a standard time format (such as ISO 8601). The format of each record is<t,ID,Value> , where Value represents the currently collected operating status value (such as temperature value, humidity value, etc.). For example, if the sensor with device ID A001 collects a temperature value of 25°C at 2025-01-06 10:00:00, it will be recorded as <2025-01-06 10:00:00,A001,25>. This process is automatically completed by the data acquisition module, and ensures the timing and source accuracy of the data.

[0082] Step 202: Integrate various operating status data to generate unified label data

[0083] After receiving multiple operating status data of the same device, the data is integrated according to the timestamp (t) to generate a unified tag data format D(t,ID). The integration process includes extracting all status data under the same timestamp from the database or cache (such as temperature T(t,ID), humidity H(t,ID), voltage V(t,ID), etc.), and combining these data into a complete tag data record. For example, if the device with device ID A001 collects temperature 25℃, humidity 60%, voltage 220V, current 10A at 2025-01-06 10:00:00, and generates a normal operation log and no alarm signal, the integrated tag data is:

[0084] D(2025-01-06 10:00:00,A001)={T=25,H=60,V=220,I=10,L=normal,A=none}.

[0085] After the integration is completed, the marked data is stored in a structured database to facilitate subsequent analysis and processing, while ensuring that the data format is unified, complete and easy to query.

[0086] Preferably, after the operating status data is marked by using the timestamp and the device ID of the corresponding device to obtain the marked data, the method further includes:

[0087] The marked data are grouped according to a preset collection period to obtain a plurality of data groups;

[0088] Calculate the difference coefficient between the current data group and the previous data group in sequence;

[0089] Determine whether the value of the coefficient of difference is within a preset range;

[0090] If the value of the difference coefficient is not within the preset range, the corresponding data group is removed;

[0091] If the value of the difference coefficient is within the preset range, the corresponding data group will be retained until all data groups are traversed to obtain the labeled data after data cleaning.

[0092] Preferably, the coefficient of difference calculation formula is:

[0093]

[0094] Among them, p X,Y is the coefficient of difference, cov(X,Y) represents the covariance between the current data set X and the previous data set Y, α X represents the mean of the current data set X, β Y Represents the mean of the previous data set Y.

[0095] Specifically, since the data acquisition device may be affected by its own parameters or environmental factors, the measurement value collected by the sensor at a certain moment may deviate greatly from the actual value. Therefore, the present application can screen out abnormal monitoring values ​​through the coefficient of difference to ensure the accuracy of the data.

[0096] Preferably, the labeled data is analyzed using a machine learning algorithm to predict the health status of the equipment and obtain a fault risk prediction result, including:

[0097] Obtain a preset sample labeled data set;

[0098] Build an initial deep learning model;

[0099] Training the initial deep learning model according to the sample labeled data set to obtain a trained classifier;

[0100] The trained LSTM neural network is connected after the classifier to obtain the fault risk prediction result.

[0101] Specifically, step 300 of this embodiment includes:

[0102] Step 301: Obtain a preset sample labeled data set

[0103] First, the historical operating status data of the equipment is collected as a sample labeled data set. The sample data includes the temperature signal, humidity signal, voltage signal, current signal, operation log and alarm signal of the equipment (i.e., data in D(t,ID) format), and the equipment health status corresponding to each data is marked (such as "normal", "slightly abnormal", "severely abnormal" or specific fault type). These annotations can be generated through the actual operation records, maintenance logs or expert experience of the equipment. In order to improve the generalization ability of the model, the sample data needs to cover a variety of equipment types, different operating environments and a variety of failure modes. At the same time, the data is cleaned and standardized to remove outliers and noise to ensure data quality. Finally, the sample data set is divided into training set, validation set and test set for model training and evaluation.

[0104] Step 302: Build an initial deep learning model

[0105] According to the characteristics of the sample labeled data, an initial deep learning model is constructed to classify the health status of the equipment. The input of the model is the multidimensional features in the labeled data (such as temperature, humidity, voltage, etc.), and the output is the health status category of the equipment. The model structure can adopt a multi-layer fully connected neural network (MLP), including an input layer, several hidden layers and an output layer. The number of nodes in the input layer is consistent with the characteristic dimension of the labeled data. The hidden layer uses the ReLU activation function to enhance the nonlinear expression ability, and the output layer uses the Softmax activation function to output the probability distribution of the health status. The loss function of the model selects the cross-entropy loss function (Cross-Entropy Loss), and the optimization algorithm uses the Adam optimizer. The model is preliminarily trained through the training set so that it can classify the health status of the equipment.

[0106] Step 303: Train the classifier and connect the LSTM neural network

[0107] After the initial deep learning model training is completed, a trained classifier is obtained to classify the current health status of the equipment. Next, the output of the classifier is used as the input of the time series data and connected to a trained LSTM (Long Short-Term Memory Network) neural network. The LSTM network is used to process the time series data of the equipment's operating status and can capture the dynamic changes and long-term dependencies of the equipment's operating status. The input of the LSTM is the health status classification results of the equipment at multiple time points and its corresponding operating status characteristics, and the output is the failure risk prediction result of the equipment (i.e., the probability of failure in the future). By training the time series sample data, the parameters of the LSTM network are optimized so that it can accurately predict the failure risk of the equipment.

[0108] Step 304: Generate failure risk prediction results

[0109] In actual applications, the real-time collected labeled data is input into the trained classifier to obtain the classification results of the current health status of the equipment. Then, these classification results and their corresponding time series data are input into the trained LSTM neural network to generate the fault risk prediction results of the equipment. The fault risk prediction results are output in the form of probability, indicating the possibility of the equipment failing in the future. For example, the fault risk prediction result of a certain device is P_fault=0.85, which means that the device has an 85% probability of failing in the future. Based on the prediction results, health scores can be assigned to the equipment, and inspection or maintenance tasks for high-risk equipment can be prioritized, thereby realizing intelligent prediction and management of the health status of the equipment.

[0110] Preferably, the calculation formula of the health score is:

[0111] HS=(1-P_fault)×100-S_abnormal×W_device;

[0112] Among them, HS is the health score, P_fault is the fault risk prediction result, S_abnormal is the abnormality of the current operating state, and W_device is the importance weight of the device; the calculation formula of the abnormality of the current operating state is: S_abnormal=Σ(|X_i-X_normal| / R_i)×W_i; Xi is the operating state data currently collected by the device, X_normal is the normal value of the operating state data corresponding to the device, R_i is the normal range of the operating state data, and W_i is the weight of the operating state data.

[0113] Specifically, the importance weight of the device (W_device) is mainly determined according to the business criticality and operation priority of the device in the computer room. Generally, core devices (such as main servers, network switches, etc.) have a greater impact on the overall operation of the computer room, so they are given a higher weight value; while auxiliary devices (such as backup servers, environmental monitoring equipment, etc.) have a smaller impact on the business and are given a lower weight value. The weight value can be determined by expert experience or business demand analysis, for example, the equipment is divided into "core equipment" (weight 5), "important equipment" (weight 3), "ordinary equipment" (weight 1) and other levels. In addition, the weight can be dynamically adjusted based on factors such as the historical failure rate and maintenance cost of the equipment. For example, if a certain device has a high historical failure rate and expensive maintenance cost, its weight value can be appropriately increased to prioritize inspection and maintenance. The weight of the operating status data (W_i) is used to measure the degree of influence of different operating parameters (such as temperature, humidity, voltage, etc.) on the health status of the equipment. The determination of the weight can be based on the technical specifications, operating experience and historical data analysis of the equipment. For example, for server equipment, the stability of voltage and current is crucial to the normal operation of the equipment, so the weight of voltage and current signals should be higher; while the influence of humidity signal is relatively small, so it is given a lower weight value. The weight value can also be determined by data-driven methods, such as using historical operation data and fault records to analyze the correlation between each operating parameter and equipment failure. The parameters with higher correlation are given higher weights. In addition, the weight value can be adjusted according to the type of equipment. For example, for air-conditioning equipment, the weight of temperature and humidity signals should be higher than that of voltage signals to reflect the operating characteristics of different equipment.

[0114] Preferably, a list of devices to be inspected is generated based on the preset inspection data and health scores of the devices, including:

[0115] Obtain the preset inspection data; the preset inspection data includes: device ID, device name, device type, device location, inspection cycle and last inspection time;

[0116] The time interval from the last inspection is calculated according to the formula T_interval=T_now-T_last; wherein T_interval is the time interval; T_now is the current time; and T_last is the last inspection time;

[0117] The inspection urgency is calculated according to the formula U_inspect=T_interval / C_device; wherein U_inspect is the inspection urgency; C_device is the inspection period;

[0118] The inspection priority of the equipment is determined according to the formula P_inspect = W_health × (1-HS / 100) + W_time × U_inspect; W_health is the weight of the health score, W_time is the weight of the inspection urgency, and P_inspect is the inspection priority;

[0119] Determine the inspection priority threshold P_threshold according to the preset inspection resources and time limit, and add the devices with inspection priority P_inspect ≥ P_threshold to the list of devices to be inspected;

[0120] Sort the devices from high to low according to the inspection priority P_inspect, give priority to high-priority devices for inspection, and generate a final sorted list of devices to be inspected.

[0121] Assume that the following equipment is in the computer room:

[0122]

[0123] Calculation steps:

[0124] Calculate the time interval:

[0125] A001: T_interval = 2025-01-06-2024-12-01 = 36 days

[0126] A002: T_interval = 2025-01-06-2024-12-25 = 12 days

[0127] A003: T_interval = 2025-01-06-2024-12-20 = 17 days

[0128] Calculate inspection urgency:

[0129] A001: U_inspect=36 / 30=1.2

[0130] A002: U_inspect=12 / 15=0.8

[0131] A003: U_inspect=17 / 20=0.85

[0132] Calculate inspection priority:

[0133] A001: P_inspect=0.6×(1-70 / 100)+0.4×1.2=0.18+0.48=0.66

[0134] A002: P_inspect=0.6×(1-50 / 100)+0.4×0.8=0.3+0.32=0.62

[0135] A003: P_inspect=0.6×(1-90 / 100)+0.4×0.85=0.06+0.34=0.4

[0136] Filter equipment to be inspected:

[0137] Set P_threshold = 0.5, and select A001 and A002 as devices to be inspected.

[0138] result:

[0139] The list of equipment to be inspected is:

[0140] Device ID health score inspection priority

[0141] A001 70 0.66

[0142] A002 50 0.62

[0143] Through the above steps, a list of equipment to be inspected can be efficiently generated, and the inspection plan can be dynamically adjusted to ensure the stable operation of the equipment in the computer room.

[0144] Optionally, the health score weight (W_health) is used to measure the impact of the equipment health status on the inspection priority. Its value depends mainly on the impact of the equipment health status on the overall operation safety of the computer room. Generally speaking, the lower the health score of the equipment, the higher the risk of failure, so a higher weight is required to prioritize the inspection. The value of W_health can be determined by the following methods:

[0145] Expert experience method: Based on the operating characteristics of the equipment and the impact of failures, the computer room operation and maintenance experts set a reasonable weight range (such as 0.5 to 0.7) to ensure that the health score plays a dominant role in the inspection priority.

[0146] Data-driven method: By analyzing historical inspection data and fault records, observing the relationship between health scores and equipment failure rates, and using regression analysis or correlation analysis to determine appropriate weight values. For example, if the health score contributes more to fault prediction, a higher weight can be assigned.

[0147] Dynamic adjustment method: In actual applications, the value of W_health is dynamically adjusted based on the inspection result feedback. For example, if a device with a low health score does have more problems during the inspection, its weight value can be appropriately increased to further optimize the inspection priority calculation.

[0148] Inspection urgency weight (W_time) is used to measure the impact of the time interval from the equipment to the preset inspection cycle on the inspection priority. Its value is mainly determined by the impact of the inspection cycle on the reliability of equipment operation. W_time can be determined in the following ways:

[0149] Matching device type with inspection cycle: Determine the weight based on the importance of the inspection cycle of the device. For example, for devices that require frequent inspections (such as UPS power supplies, air conditioners, etc.), the inspection cycle is more important, so a higher W_time value (such as 0.4-0.6) is assigned; while for devices with longer inspection cycles (such as standby servers), a lower weight value is assigned.

[0150] Historical data analysis method: Analyze historical inspection records and observe the change in the probability of equipment failure after the inspection period has exceeded. If it is found that the inspection period has a greater impact on the equipment failure rate, the weight value of W_time should be increased.

[0151] Balancing method: In order to achieve a reasonable balance between health score and inspection urgency, the ratio of W_health and W_time can be adjusted through experiments (such as W_health + W_time = 1) to ensure that the contribution of both to the inspection priority meets actual needs. For example, for scenarios with large fluctuations in health scores, the weight of W_time can be appropriately reduced to highlight the impact of health scores.

[0152] Through the above method, the weights of W_health and W_time can be dynamically adjusted to meet the actual conditions of different device types, operating environments, and inspection requirements.

[0153] Preferably, the path optimization algorithm is Dijkstra algorithm.

[0154] Specifically, step 500 of this embodiment includes:

[0155] Step 501: Build a graph model of device location

[0156] First, a graph model is constructed based on the geographic location of each device in the device list. The nodes of the graph represent the location of the device, including the inspection starting point (such as the entrance to the machine room) and the location of all devices to be inspected; the edges of the graph represent the paths between devices, and the weight of the edge is the distance between the devices or the time required for the inspection. The distance can be calculated by the geographic coordinates of the device (such as Euclidean distance) or measured based on the actual layout of the machine room (such as floors, passages). If there are obstacles or restrictions on the path between devices (such as elevators between floors, closed areas, etc.), these restrictions need to be marked in the graph to ensure the feasibility of the path.

[0157] Step 502: Initialize the parameters of Dijkstra algorithm

[0158] After building the graph model, initialize the parameters of the Dijkstra algorithm. Select the inspection starting point as the source node, set the distance from the starting point to itself to 0, and set the initial distances of the remaining nodes to infinity (∞). At the same time, create a priority queue to store unvisited nodes and record the shortest path distance and predecessor node (for backtracking) of each node. The priority queue is sorted based on the shortest path distance of the current node, ensuring that the node with the smallest distance is taken out of the queue for expansion each time.

[0159] Step 503: Execute Dijkstra algorithm to calculate the shortest path

[0160] Starting from the starting point, take the node with the shortest distance from the priority queue as the current node, and update the shortest path distance of its neighbor nodes. For each neighbor node, calculate the distance from the starting point through the current node to the neighbor node, and compare it with the shortest distance currently recorded:

[0161] distance[neighbor]=min(distance[neighbor],distance[current]+weight[current][neighbor])

[0162] If the new distance is shorter, the shortest distance of the neighbor node is updated and added to the priority queue. This process is repeated until all nodes are visited or the shortest path to the target node is determined. Eventually, the algorithm generates a shortest path tree from the starting point to all device nodes.

[0163] Step 504: Generate an optimal inspection path

[0164] According to the calculation results of the Dijkstra algorithm, starting from the starting point, all devices to be inspected are visited in sequence to generate the optimal inspection path. The complete path from the starting point to the target node can be constructed by backtracking the predecessor node of each node. If the inspection task needs to visit all devices and return to the starting point (i.e., form a closed-loop path), other algorithms (such as greedy algorithms or dynamic programming) can be combined on the basis of the optimal path to further optimize the path sequence and ensure the shortest total path length.

[0165] Step 505: Output the optimal path and apply it to the inspection task

[0166] The generated optimal inspection path is output as a specific inspection order list, including the inspection starting point and the access order of all devices. For example, the path can be expressed as: S→A→B→C→D→S, where S is the starting point and A, B, C, and D are the devices to be inspected. The path information can be displayed to the inspectors through visualization tools (such as the floor plan of the computer room) and sent to the inspection terminal device (such as a mobile phone or tablet). The inspectors complete the equipment inspection in sequence according to the path to ensure the efficiency and accuracy of the inspection. If the equipment status changes during the inspection (such as the addition of a high-priority device), the Dijkstra algorithm can be rerun in real time to generate a new optimal path.

[0167] Corresponding to the above method, such as Figure 2 As shown, this embodiment provides an intelligent distribution system for sequential inspection of a computer room, including:

[0168] A data acquisition unit, used to deploy IoT sensors and monitoring equipment in the computer room to collect operating status data of the equipment inside the computer room in real time;

[0169] A data marking unit, used to mark the operation status data with a timestamp and a device ID of a corresponding device to obtain marked data;

[0170] A scoring calculation unit, configured to analyze the labeled data using a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assign a health score to each equipment according to the fault risk prediction result;

[0171] A list generation unit, used to generate a list of devices to be inspected according to a preset inspection cycle and health score of the devices;

[0172] A path optimization unit, configured to generate an optimal inspection path according to the geographical location of each device in the device list based on a path optimization algorithm;

[0173] The sequential inspection unit is used to send the optimal inspection path to the inspection personnel to realize the intelligent allocation of the sequential inspection of the computer room.

[0174] The beneficial effects of the present invention are as follows:

[0175] (1) The present invention collects equipment operation status data in real time through Internet of Things sensors and generates an optimal inspection path in combination with a path optimization algorithm, which can significantly reduce the inspection time and unnecessary repetitive work of inspection personnel and improve inspection efficiency.

[0176] (2) The present invention uses machine learning algorithms to analyze labeled data, predict the health status of equipment and generate fault risk prediction results. A health score can be assigned to each device, thereby helping managers to understand the operating status of the equipment more accurately and discover potential problems in a timely manner.

[0177] (3) The present invention dynamically generates a list of devices to be inspected based on the health scores of the devices and the preset inspection cycles, which can reasonably arrange inspection tasks, avoid waste of resources, and ensure that key equipment receives priority inspection.

[0178] (4) The present invention can detect potential problems in advance and take preventive measures through real-time monitoring of the equipment operating status and prediction of failure risks, thereby reducing the equipment failure rate and ensuring the normal operation of the computer room.

[0179] (5) The present invention can automatically complete data collection, analysis, task allocation and path planning, reducing the need for human intervention and the risk of errors caused by human factors.

[0180] (6) Inspection personnel can efficiently complete their tasks by receiving the optimal inspection route, reducing workload and unnecessary complexity, and improving the inspection work experience and satisfaction.

[0181] (7) Since the present invention can collect data in real time and generate a dynamic inspection plan, it can quickly respond to changes in equipment status, adapt to the real-time needs of the computer room environment, and improve the flexibility and intelligence level of inspection.

[0182] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0183] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. An intelligent allocation method for sequential inspection of a computer room, characterized in that: include: Deploy IoT sensors and monitoring equipment in the computer room to collect real-time operating status data of the equipment inside the computer room; Marking the operation status data using a timestamp and a device ID of a corresponding device to obtain marked data; Analyzing the labeled data using a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assigning a health score to each device according to the fault risk prediction result; Generate a list of devices to be inspected based on the preset inspection cycle and health score of the devices; Based on the path optimization algorithm, an optimal inspection path is generated according to the geographical location of each device in the device list; The optimal inspection path is sent to the inspection personnel to realize the intelligent allocation of sequential inspection of the computer room.

2. The intelligent allocation method for sequential inspection of a computer room according to claim 1 is characterized in that: The operation status data includes: temperature signal, humidity signal, voltage signal, current signal, equipment operation log and alarm signal.

3. The intelligent allocation method for sequential inspection of a computer room according to claim 2 is characterized in that: The operation status data is marked using a timestamp and a device ID of a corresponding device to obtain marked data, including: Add a timestamp and device ID to each collected operation status data record; the timestamp is used to record the specific time point of data collection, and the device ID is used to identify the specific device from which the data comes; Integrate multiple operating status data of the same device according to timestamps to generate unified tag data; the format of the tag data is: D(t,ID)={T(t,ID),H(t,ID),V(t,ID),I(t,ID),L(t,ID),A(t,ID)}; wherein D(t,ID) represents a set of tag data collected by the device ID at time t; T(t,ID) represents the temperature signal value of the device ID at time t; H(t,ID) represents the humidity signal value of the device ID at time t; V(t,ID) represents the voltage signal value of the device ID at time t; I(t,ID) represents the current signal value of the device ID at time t; L(t,ID) represents the operating log of the device ID at time t; A(t,ID) represents the alarm signal of the device ID at time t; t represents the specific time point of data collection; ID represents the unique identifier of the device.

4. The intelligent allocation method for sequential inspection of a computer room according to claim 1 is characterized in that: After the operation status data is marked by using the timestamp and the device ID of the corresponding device to obtain the marked data, the method further includes: The marked data are grouped according to a preset collection period to obtain a plurality of data groups; Calculate the difference coefficient between the current data group and the previous data group in sequence; Determine whether the value of the coefficient of difference is within a preset range; If the value of the difference coefficient is not within the preset range, the corresponding data group is removed; If the value of the difference coefficient is within the preset range, the corresponding data group will be retained until all data groups are traversed to obtain the labeled data after data cleaning.

5. The intelligent allocation method for sequential inspection of a computer room according to claim 4 is characterized in that: The coefficient of difference calculation formula is: Among them, p X,Y is the coefficient of difference, cov(X,Y) represents the covariance between the current data set X and the previous data set Y, α X represents the mean of the current data set X, β Y Represents the mean of the previous data set Y.

6. The intelligent allocation method for sequential inspection of a computer room according to claim 1 is characterized in that: The labeled data is analyzed using a machine learning algorithm to predict the health status of the equipment and obtain a fault risk prediction result, including: Obtain a preset sample labeled data set; Build an initial deep learning model; Training the initial deep learning model according to the sample labeled data set to obtain a trained classifier; The trained LSTM neural network is connected after the classifier to obtain the fault risk prediction result.

7. The intelligent allocation method for sequential inspection of a computer room according to claim 1 is characterized in that: The calculation formula of the health score is: HS=(1-P_fault)×100-S_abnormal×W_device; Among them, HS is the health score, P_fault is the fault risk prediction result, S_abnormal is the abnormality of the current operating state, and W_device is the importance weight of the device; the calculation formula of the abnormality of the current operating state is: S_abnormal=Σ(|X_i-X_normal| / R_i)×W_i; Xi is the operating state data currently collected by the device, X_normal is the normal value of the operating state data corresponding to the device, R_i is the normal range of the operating state data, and W_i is the weight of the operating state data.

8. The intelligent allocation method for sequential inspection of a computer room according to claim 7 is characterized in that: Generate a list of devices to be inspected based on the preset inspection data and health scores of the devices, including: Obtain the preset inspection data; the preset inspection data includes: device ID, device name, device type, device location, inspection cycle and last inspection time; The time interval from the last inspection is calculated according to the formula T_interval=T_now-T_last; wherein T_interval is the time interval; T_now is the current time; and T_last is the last inspection time; The inspection urgency is calculated according to the formula U_inspect=T_interval / C_device; wherein U_inspect is the inspection urgency; C_device is the inspection period; The inspection priority of the equipment is determined according to the formula P_inspect = W_health × (1-HS / 100) + W_time × U_inspect; W_health is the weight of the health score, W_time is the weight of the inspection urgency, and P_inspect is the inspection priority; Determine the inspection priority threshold P_threshold according to the preset inspection resources and time limit, and add the devices with inspection priority P_inspect ≥ P_threshold to the list of devices to be inspected; Sort the devices from high to low according to the inspection priority P_inspect, give priority to high-priority devices for inspection, and generate a final sorted list of devices to be inspected.

9. The intelligent allocation method for sequential inspection of a computer room according to claim 1, characterized in that: The path optimization algorithm is Dijkstra algorithm.

10. An intelligent distribution system for sequential inspection of a machine room, characterized in that: include: A data acquisition unit, used to deploy IoT sensors and monitoring equipment in the computer room to collect operating status data of the equipment inside the computer room in real time; A data marking unit, used to mark the operation status data with a timestamp and a device ID of a corresponding device to obtain marked data; A scoring calculation unit, configured to analyze the labeled data using a machine learning algorithm to predict the health status of the equipment, obtain a fault risk prediction result, and assign a health score to each equipment according to the fault risk prediction result; A list generation unit, used to generate a list of devices to be inspected according to a preset inspection cycle and health score of the devices; A path optimization unit, configured to generate an optimal inspection path according to the geographical location of each device in the device list based on a path optimization algorithm; The sequential inspection unit is used to send the optimal inspection path to the inspection personnel to realize the intelligent allocation of the sequential inspection of the computer room.

Images