Network equipment diagnosis and control method based on artificial intelligence

Abstract

The invention discloses a network equipment diagnosis and control method based on artificial intelligence, and relates to the technical field of equipment diagnosis and control. Setting an inspection period, regularly obtaining the business volume and the data interaction rate of the network equipment according to the inspection period, and judging whether the network equipment is abnormal or not according to the business volume and the data interaction rate; if no abnormity exists, no operation needs to be carried out; if yes, acquiring a three-dimensional conventional thermal inductance model and a three-dimensional real-time thermal inductance model of the network equipment, and checking whether the three-dimensional real-time thermal inductance model has an abnormal part or not through the three-dimensional conventional thermal inductance model; if the abnormal part exists, the network equipment has an offline fault, corresponding offline fault information is generated, and corresponding measures are taken according to the offline fault information; if the abnormal part does not exist, the network equipment does not have offline faults, early warning information is generated, and corresponding measures are taken according to the early warning information.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment diagnosis and control, and particularly to a method for diagnosing and controlling network equipment based on artificial intelligence. Background Art

[0002] With the continuous development of network technology and the increasing number of Internet devices, the stability and reliability of network equipment have become the key to maintaining and managing network operations. Traditional network equipment diagnosis methods mostly rely on manual inspections and simple fault detection tools. Although some fault problems can be found, the ability to predict complex equipment faults and performance anomalies is limited. Especially when equipment fails, there is a lack of efficient automated diagnosis means, which easily leads to an extended equipment downtime and thus affects the normal operation of the network.

[0003] In recent years, with the rapid development of artificial intelligence (AI) technology, intelligent diagnosis methods have gradually made remarkable progress in various fields. Especially in network equipment diagnosis, AI technology can achieve anomaly detection and fault prediction of network equipment through methods such as deep learning, image recognition, and data mining. However, existing network equipment fault diagnosis technologies still have certain limitations, such as being difficult to effectively process multi-dimensional and complex data, and lacking comprehensive equipment status monitoring means, resulting in the inability to timely and accurately discover potential faults or performance degradation problems.

[0004] How to real-time monitor and detect the operating status of equipment, and perform automated diagnosis and processing on equipment anomalies, how to accurately determine whether there are anomalies in the equipment, and then take corresponding diagnostic measures to improve the efficiency of equipment fault handling, reduce manual intervention, and achieve dynamic monitoring of the equipment health status is the problem we need to solve. For this reason, a method for diagnosing and controlling network equipment based on artificial intelligence is provided. Summary of the Invention

[0005] In order to solve the above problems, the purpose of the present invention is to provide a method for diagnosing and controlling network equipment based on artificial intelligence.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A method for diagnosing and controlling network equipment based on artificial intelligence, including the following steps:

[0007] Step S1: Set an inspection period, regularly obtain the traffic volume and data interaction rate of the network equipment according to the inspection period, and determine whether there are anomalies in the network equipment according to the traffic volume and data interaction rate;

[0008] Step S2: If there are no anomalies, no operation is required; if there are anomalies, obtain the three-dimensional conventional thermal sense model and three-dimensional real-time thermal sense model of the network equipment, and check whether there are abnormal parts in the three-dimensional real-time thermal sense model through the three-dimensional conventional thermal sense model;

[0009] Step S3: If there is an abnormal part, the network device has an offline fault, generate corresponding offline fault information, and take corresponding measures according to the offline fault information; if there is no abnormal part, the network device has no offline fault, generate a warning message, and take corresponding measures according to the warning message.

[0010] Further, set an inspection period, regularly obtain the traffic volume and data interaction rate of the network device according to the inspection period, and the process of judging whether the network device is abnormal according to the traffic volume and data interaction rate includes:

[0011] The inspection period includes a peak inspection period and a trough inspection period;

[0012] Obtain the number of concurrent connections and requests of the network device, and generate the traffic volume of the network device according to the number of concurrent connections and requests; obtain the throughput, bandwidth and delay of the network device, and generate the data interaction rate of the network device according to the throughput, bandwidth and delay.

[0013] Set a periodic detection time period;

[0014] Obtain the traffic volume and data interaction rate of the network device within the periodic detection time period;

[0015] Establish a three-dimensional coordinate system A, the three-dimensional coordinate system A includes an X A axis, a Y A axis and a Z A axis, generate a two-dimensional coordinate system A-XY according to the X A axis and the Y A axis, and generate a two-dimensional coordinate system A-XZ according to the X A axis and the Z A axis;

[0016] According to the acquisition time of the traffic volume, map the traffic volume to the two-dimensional coordinate system A-XY to generate a traffic volume curve;

[0017] According to the acquisition time of the data interaction rate, map the data interaction rate to the two-dimensional coordinate system A-XZ to generate a data interaction rate curve;

[0018] Generate a number of periodic detection time nodes according to the periodic detection time period;

[0019] Obtain the slope of the points corresponding to each periodic detection time node in the traffic volume curve, and record the slope as the traffic conversion rate; obtain the slope of the points corresponding to each periodic detection time node in the data interaction rate curve, and record the slope as the interaction conversion rate;

[0020] Set a traffic threshold and an interaction threshold;

[0021] Obtain the cycle detection time nodes corresponding to the service conversion rate ≥ the service threshold, and mark the cycle detection time nodes as service peak time nodes;

[0022] Obtain the cycle detection time nodes corresponding to the interaction conversion rate ≥ the interaction threshold, and mark the cycle detection time nodes as interaction peak time nodes;

[0023] Generate a peak inspection period based on the overlapping time nodes between the service peak time nodes and the interaction peak time nodes, and generate a trough inspection period based on the time nodes other than the corresponding time nodes of the peak inspection period;

[0024] Set the peak normal index range and the trough normal index range;

[0025] Regularly obtain the service volume and data interaction rate of the network device according to the inspection period, and generate a status index based on the service volume and the data interaction rate;

[0026] If the status index generated within the peak inspection period ∈ the peak normal index range, the network device has no abnormality, otherwise, there is an abnormality;

[0027] If the status index generated within the trough inspection period ∈ the trough normal index range, the network device has no abnormality, otherwise, there is an abnormality.

[0028] Further, the process of obtaining the three-dimensional normal thermal sense model and the three-dimensional real-time thermal sense model of the network device includes:

[0029] Step a1: Set the sampling period; Set a number of infrared thermal imagers around the network device, and obtain the infrared thermal imaging video of the network device under normal conditions through the infrared thermal imagers according to the sampling period, and mark the infrared thermal imaging video as the normal thermal imaging video;

[0030] Step a2: Divide the normal thermal imaging video into several images in units of frames, mark the images as normal thermal images, and then generate a normal thermal image set from the normal thermal images generated by different infrared thermal imagers at the same time;

[0031] Step a3: Establish a three-dimensional solid model of the network device;

[0032] Step a4: Attach the normal thermal images in the normal thermal image set to the three-dimensional solid model of the network device in chronological order to generate a number of three-dimensional thermal sense sub-models; that is, each of the number of three-dimensional thermal sense sub-models has its corresponding time, and the time is the time corresponding to the normal thermal image corresponding to the three-dimensional thermal sense sub-model;

[0033] Step a5: Establish a four-dimensional coordinate system, which includes an X-axis, a Y-axis, a Z-axis, and a time axis. Map a number of three-dimensional thermal sensing sub-models to the four-dimensional coordinate system in chronological order, and generate a three-dimensional conventional thermal sensing model through a three-dimensional convolutional neural network;

[0034] According to the principle of obtaining the three-dimensional normal thermal sensing model of the network device in steps a1 to a5, obtain the three-dimensional real-time thermal sensing model of the network device.

[0035] Further, the process of using the three-dimensional conventional thermal sensing model to check whether there are abnormal parts in the three-dimensional real-time thermal sensing model includes:

[0036] Set a normal sensing interval;

[0037] Obtain the coordinates corresponding to the RGB values ∈ normal sensing interval in the three-dimensional conventional thermal sensing model, and generate a normal sensing area according to the coordinates;

[0038] Obtain the maximum and minimum values of the normal sensing area on the X-axis, Y-axis, and Z-axis according to the four-dimensional coordinate system, and generate the center coordinates of the normal sensing area according to the maximum and minimum values;

[0039] Taking the center coordinates as the center and the X-axis, Y-axis, and Z-axis as the value-taking directions, obtain the boundary numerical change set of the normal sensing area according to the time axis;

[0040] Obtain the X-value variation range, Y-value variation range, Z-value variation range, X-value variation rate, Y-value variation rate, and Z-value variation rate according to the boundary numerical change set;

[0041] Generate a boundary change feature set according to the X-value variation range, Y-value variation range, Z-value variation range, X-value variation rate, Y-value variation rate, and Z-value variation rate;

[0042] Obtain the coordinates corresponding to the RGB values ∈ normal sensing interval in the three-dimensional real-time thermal sensing model, and generate a real sensing area according to the coordinates;

[0043] Obtain the center coordinates in the three-dimensional conventional thermal sensing model, check whether all the center coordinates are within the real sensing area and whether there are center coordinates within the real sensing area. If all the center coordinates are within the real sensing area and there are center coordinates within the real sensing area, perform a secondary module check, and determine whether there are abnormal parts in the three-dimensional real-time thermal sensing model through the secondary module check; otherwise, there are abnormal parts. If there are center coordinates not within the real sensing area, generate abnormal position information according to the center coordinates. If there are no center coordinates within the real sensing area, generate abnormal position information according to the real sensing area.

[0044] Further, the process of performing a secondary module check and determining whether there are abnormal parts in the three-dimensional real-time thermal sensing model through the secondary module check includes:

[0045] According to the principle process of obtaining the boundary change feature set corresponding to the normal sensing area based on the sensed heart coordinates, obtain the boundary change feature set corresponding to the actual sensing area based on the sensed heart coordinates;

[0046] If the element ranges in the boundary change feature set corresponding to the actual sensing area based on the sensed heart coordinates are all within the ranges of the corresponding elements in the boundary change feature set corresponding to the normal sensing area based on the sensed heart coordinates, it is determined that there is no abnormal part in the three-dimensional real-time thermal sensing model; otherwise, it is determined that there is an abnormal part in the three-dimensional real-time thermal sensing model;

[0047] Obtain the sensed heart coordinates corresponding to the determination that there is an abnormal part in the three-dimensional real-time thermal sensing model, and generate abnormal position information according to the sensed heart coordinates.

[0048] Furthermore, the process of generating corresponding offline fault information and taking corresponding measures according to the offline fault information includes:

[0049] Generate offline fault information according to the abnormal position information, and send the offline fault information to the user to inform the user of the position information where an abnormality is suspected, and let the user perform corresponding processing.

[0050] Furthermore, the process of taking corresponding measures according to the warning information includes:

[0051] Send the warning information to the user to inform the user that there is no offline fault and an online fault is suspected, and let the user take corresponding measures.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] The present invention generates a three-dimensional thermal sensing model through artificial intelligence, obtains and analyzes the operating state of network devices in real time; when an abnormality occurs in the device, it can accurately identify the fault part through the thermal sensing model, and classify and process according to the abnormal information, improving the accuracy and response speed of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0055] As Figure 1 shown, the network device diagnosis and control method based on artificial intelligence includes the following steps:

[0056] Step S1: Set an inspection period, regularly obtain the traffic volume and data interaction rate of the network device according to the inspection period, and determine whether there is an abnormality in the network device according to the traffic volume and data interaction rate;

[0057] Step S2: If there is no anomaly, no operation is required; if there is an anomaly, obtain the three-dimensional conventional thermal model and the three-dimensional real-time thermal model of the network device, and check whether there are any abnormal parts in the three-dimensional real-time thermal model through the three-dimensional conventional thermal model;

[0058] Step S3: If there are abnormal parts, the network device has an offline fault, generate the corresponding offline fault information, and take corresponding measures according to the offline fault information; if there are no abnormal parts, the network device has no offline fault, generate a warning message, and take corresponding measures according to the warning message.

[0059] It should be further noted that in the specific implementation process, set an inspection period, regularly obtain the traffic volume and data interaction rate of the network device according to the inspection period, and the process of judging whether the network device has an anomaly based on the traffic volume and data interaction rate includes:

[0060] The inspection period includes a peak inspection period and a trough inspection period;

[0061] Obtain the concurrent connection number and request number of the network device, and generate the traffic volume of the network device according to the concurrent connection number and request number;

[0062] Set the traffic coefficient τ 1 and the traffic coefficient τ 2 ;

[0063] The traffic volume = τ 1 × concurrent connection number + τ 2 × request number;

[0064] Obtain the throughput, bandwidth, and latency of the network device, and generate the data interaction rate of the network device according to the throughput, bandwidth, and latency;

[0065] Set the interaction coefficient ω 1 、the interaction coefficient ω 2 and the interaction coefficient ω 3 ;

[0066] The data interaction rate = ω 1 × throughput + ω 2 × bandwidth + ω 3 × latency;

[0067] Set the period detection time period;

[0068] Obtain the traffic volume and data interaction rate of the network device within the period detection time period;

[0069] Establish a three-dimensional coordinate system A, and the three-dimensional coordinate system A includes the X A axis, the Y A axis and the Z A axis, and according to the X AAxis and Y A The X-axis and the Y-axis generate a two-dimensional coordinate system A-XY. According to the X A axis and the Z A axis generate a two-dimensional coordinate system A-XZ;

[0070] According to the acquisition time of the business volume, map the business volume to the two-dimensional coordinate system A-XY to generate a business volume curve;

[0071] According to the acquisition time of the data interaction rate, map the data interaction rate to the two-dimensional coordinate system A-XZ to generate a data interaction rate curve;

[0072] Generate a number of periodic detection time nodes according to the periodic detection period;

[0073] Obtain the slope of the points corresponding to each periodic detection time node in the business volume curve, and denote the slope as the business conversion rate;

[0074] Obtain the slope of the points corresponding to each periodic detection time node in the data interaction rate curve, and denote the slope as the interaction conversion rate;

[0075] Set a business threshold and an interaction threshold;

[0076] Obtain the periodic detection time nodes corresponding to the business conversion rate ≥ the business threshold, and mark the periodic detection time nodes as business peak time nodes;

[0077] Obtain the periodic detection time nodes corresponding to the interaction conversion rate ≥ the interaction threshold, and mark the periodic detection time nodes as interaction peak time nodes;

[0078] Generate a peak inspection period according to the overlapping time nodes between the business peak time nodes and the interaction peak time nodes, and generate a trough inspection period according to the time nodes other than the time nodes corresponding to the peak inspection period;

[0079] It should be noted that the time intervals for obtaining the business volume and the data interaction rate are different within different inspection periods;

[0080] Set a peak normal index interval and a trough normal index interval;

[0081] Regularly obtain the business volume and the data interaction rate of the network device according to the inspection period, and generate a status index according to the business volume and the data interaction rate;

[0082]

[0083] If the status index generated within the peak inspection period ∈ the peak normal index interval, the network device has no abnormality, otherwise, there is an abnormality;

[0084] If the status index generated within the low valley inspection period belongs to the low valley normal index range, then there is no abnormality in the network device; otherwise, there is an abnormality.

[0085] It should be further noted that in the specific implementation process, the process of obtaining the three-dimensional normal thermal sense model and the three-dimensional real-time thermal sense model of the network device includes:

[0086] Step a1: Set the sampling period; set a number of infrared thermal imagers around the network device, and obtain the infrared thermal imaging video of the network device in the normal state through the infrared thermal imagers according to the sampling period, and mark the infrared thermal imaging video as the normal thermal imaging video.

[0087] Step a2: Divide the normal thermal imaging video into several images in units of frames, mark the images as normal thermal images, and then generate a normal thermal image set from the normal thermal images generated by different infrared thermal imagers at the same time.

[0088] Step a3: Establish a three-dimensional solid model of the network device.

[0089] Step a4: Attach the normal thermal images in the normal thermal image set to the three-dimensional solid model of the network device in chronological order to generate several three-dimensional thermal sense sub-models; that is, each of the several three-dimensional thermal sense sub-models has its corresponding time, and the time is the time corresponding to the normal thermal image of the three-dimensional thermal sense sub-model.

[0090] Step a5: Establish a four-dimensional coordinate system, which includes the X-axis, Y-axis, Z-axis and time axis, map several three-dimensional thermal sense sub-models to the four-dimensional coordinate system in chronological order, and generate a three-dimensional normal thermal sense model through a three-dimensional convolutional neural network.

[0091] According to the principle of obtaining the three-dimensional normal thermal sense model of the network device in steps a1 to a5, obtain the three-dimensional real-time thermal sense model of the network device.

[0092] It should be noted that the infrared thermal imagers around the network device start recording and stop recording at the same time, that is, the start time and end time of the normal thermal imaging video generated by the infrared thermal imagers are the same.

[0093] The normal thermal image set 取样时间 ={normal thermal image 1 , normal thermal image 2 , ……, normal thermal image n}, where n is a positive integer, and the sampling time is the generation time of the normal thermal image in the normal thermal image set.

[0094] The attachment is to match according to the texture features of the conventional thermal image and the geometric relationship of the three-dimensional solid model, obtain the RGB values corresponding to each pixel point in the conventional thermal image, and then assign the RGB values of the pixel points in the conventional thermal image to the coordinates in the three-dimensional solid model according to the matching result to generate a three-dimensional thermal sub-model, that is, each coordinate in the three-dimensional thermal sub-model has its corresponding RGB value;

[0095] If there are multiple RGB values corresponding to the pixel points in the conventional thermal image assigned to one coordinate in the three-dimensional thermal sub-model, then take the average value of the RGB values to assign to this coordinate;

[0096] It should be further noted that in the specific implementation process, the process of using the three-dimensional conventional thermal model to check whether there are abnormal parts in the three-dimensional real-time thermal model includes:

[0097] Set the normal feeling interval;

[0098] Obtain the coordinates corresponding to the RGB values ∈ the normal feeling interval in the three-dimensional conventional thermal model, and generate a normal feeling area according to the coordinates;

[0099] According to the four-dimensional coordinate system, obtain the maximum and minimum values of the normal feeling area on the X-axis, Y-axis and Z-axis, and generate the center coordinates of the normal feeling area according to the maximum and minimum values;

[0100] The maximum value includes X max , Y max and Z max ;

[0101] The minimum value includes X min , Y min and Z min ;

[0102]

[0103] With the center coordinates as the center and the X-axis, Y-axis and Z-axis as the value-taking directions, obtain the boundary numerical change set of the normal feeling area according to the time axis;

[0104] The where n is a positive integer; is the maximum value corresponding to the normal feeling area on the X-axis with the center coordinates as the center and the X-axis as the value-taking direction when the time axis is time_i; is the minimum value corresponding to the normal feeling area on the X-axis with the center coordinates as the center and the X-axis as the value-taking direction when the time axis is time_i; is the maximum value corresponding to the normal feeling area on the Y-axis with the center coordinates as the center and the Y-axis as the value-taking direction when the time axis is time_i; When the time axis is time_i, with the center of the sensing center coordinates and the Y-axis as the value-taking direction, it is the corresponding minimum value of the constant sensing region on the Y-axis; When the time axis is time_i, with the center of the sensing center coordinates and the Z-axis as the value-taking direction, it is the corresponding maximum value of the constant sensing region on the Z-axis; When the time axis is time_i, with the center of the sensing center coordinates and the Z-axis as the value-taking direction, it is the corresponding minimum value of the constant sensing region on the Z-axis;

[0105] Obtain the X value range, Y value range, Z value range, X value change rate, Y value change rate, and Z value change rate according to the boundary value change set;

[0106] Obtain in the boundary value change set The maximum value and the minimum value, which are respectively denoted as and Obtain in the boundary value change set The maximum value and the minimum value, which are respectively denoted as and According to and Generate the X value range; the

[0107] According to the adjacent and in the boundary value change set, generate the positive ratio and the negative ratio Obtain the adjacent positive ratios and The difference between them, the maximum value and the minimum value are respectively denoted as and Obtain the adjacent negative ratios and The difference between them, the maximum value and the minimum value are respectively denoted as and According to and Generate the X value change rate; the

[0108] According to the principle of obtaining the X value range through and Obtain the Y value range through and Obtain the Z value range through and ;

[0109] According to the principle of obtaining the X value change rate through and ​ and Obtain the rate of change of the Y value, by and obtain the rate of change of the Z value;

[0110] Generate a boundary change feature set according to the X value variation range, Y value variation range, Z value variation range, X value change rate, Y value change rate, and Z value change rate;

[0111] The boundary change feature set = {(X value variation range, X value change rate), (Y value variation range, Y value change rate), (Z value variation range, Z value change rate)};

[0112] Obtain the coordinates corresponding to the RGB values ∈ the normal sense interval in the three-dimensional real-time thermal sense model, and generate a real sense area according to the coordinates;

[0113] Obtain the center sense coordinates in the three-dimensional conventional thermal sense model, check whether all the center sense coordinates are within the real sense area and whether there are center sense coordinates within the real sense area. If all the center sense coordinates are within the real sense area and there are center sense coordinates within the real sense area, then perform a secondary module check to determine whether there are abnormal parts in the three-dimensional real-time thermal sense model; otherwise, there are abnormal parts. If there are center sense coordinates not within the real sense area, then generate abnormal position information according to the center sense coordinates. If there are no center sense coordinates within the real sense area, then generate abnormal position information according to the real sense area;

[0114] Among them, the process of performing a secondary module check to determine whether there are abnormal parts in the three-dimensional real-time thermal sense model includes:

[0115] According to the principle process of obtaining the boundary change feature set corresponding to the normal sense area corresponding to the center sense coordinates, obtain the boundary change feature set corresponding to the real sense area corresponding to the center sense coordinates;

[0116] If the element ranges in the boundary change feature set corresponding to the real sense area corresponding to the center sense coordinates are all within the corresponding element ranges in the boundary change feature set corresponding to the normal sense area corresponding to the center sense coordinates, then determine that there are no abnormal parts in the three-dimensional real-time thermal sense model; otherwise, determine that there are abnormal parts in the three-dimensional real-time thermal sense model;

[0117] Obtain the center sense coordinates corresponding to the determination that there are abnormal parts in the three-dimensional real-time thermal sense model, and generate abnormal position information according to the center sense coordinates;

[0118] Assume that there is an element in the boundary change feature set corresponding to the real sense area corresponding to a certain center sense coordinate Obtain the corresponding element in the boundary change feature set corresponding to the normal sense area corresponding to the center sense coordinate If and Then, the range of the X value of the elements in the boundary change feature set corresponding to the actual feeling area of the feeling center coordinate is within the range of the corresponding elements in the boundary change feature set corresponding to the normal feeling area of the feeling center coordinate;

[0119] It should be further noted that, in the specific implementation process, the process of generating the corresponding offline fault information and taking corresponding measures according to the offline fault information includes:

[0120] Generating offline fault information according to the abnormal position information, and sending the offline fault information to the user to inform the user of the position information where an abnormality is suspected, and the user takes corresponding measures;

[0121] It should be further noted that, in the specific implementation process, the process of taking corresponding measures according to the warning information includes:

[0122] Sending the warning information to the user to inform the user that there is no offline fault and an online fault is suspected, and the user takes corresponding measures;

[0123] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A network equipment diagnosis and control method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Setting a test period, regularly obtaining the service volume and data exchange rate of the network device according to the test period, and judging whether the network device has an abnormality according to the service volume and data exchange rate; Step S2: If there is no abnormality, no operation is required; if there is an abnormality, a 3D conventional thermal model and a 3D real-time thermal model of the network device are obtained, and the 3D real-time thermal model is checked by the 3D conventional thermal model to see if there is an abnormal part in the 3D real-time thermal model; Step S3: If there is an abnormal part, then the network device has an offline fault, and corresponding offline fault information is generated, and corresponding measures are taken according to the offline fault information; If there is no abnormal part, then there is no offline fault in the network equipment, and an early warning message is generated, and corresponding measures are taken according to the early warning message.

2. The artificial intelligence-based network equipment diagnosis and control method according to claim 1, characterized in that: Set a test cycle, regularly obtain the service volume and data exchange rate of network devices according to the test cycle, and determine whether there is an abnormality in the network device based on the service volume and data exchange rate. The process includes: The inspection cycle includes a peak inspection cycle and a valley inspection cycle; Obtain the number of concurrent connections and requests of a network device, and generate the traffic volume of the network device according to the number of concurrent connections and requests; obtain the throughput, bandwidth and latency of the network device, and generate the data interaction rate of the network device according to the throughput, bandwidth and latency; Set the periodic detection period; Obtain the traffic volume and data exchange rate of network devices during the periodic detection period; Establish a three-dimensional coordinate system A, the three-dimensional coordinate system A includes X A Axis, Y A Axis and Z A Axis, according to X A Axis and Y A Axis generates a two-dimensional coordinate system A-XY, according to X A Axis and Z A The axis generates a two-dimensional coordinate system A-XZ; According to the acquisition time of the business volume, the business volume is mapped to a two-dimensional coordinate system A-XY to generate a business volume curve; According to the acquisition time of the data interaction rate, the data interaction rate is mapped to a two-dimensional coordinate system A-XZ to generate a data interaction rate curve; Generate a number of periodic detection time nodes according to the periodic detection period; Obtain the slope of the point corresponding to each periodic detection time node in the business volume curve, and record the slope as the business conversion rate; obtain the slope of the point corresponding to each periodic detection time node in the data interaction rate curve, and record the slope as the interaction conversion rate; Set business thresholds and interaction thresholds; Obtain a periodic detection time node corresponding to a service conversion rate ≥ a service threshold, and mark the periodic detection time node as a service peak time node; Obtain a periodic detection time node corresponding to an interaction conversion rate ≥ an interaction threshold, and mark the periodic detection time node as an interaction peak time node; Generate a peak inspection cycle based on the overlapping time nodes between the business peak time nodes and the interaction peak time nodes, and generate a valley inspection cycle based on the time nodes other than the corresponding time nodes of the peak inspection cycle; Set the peak regular index range and the trough regular index range; Regularly acquiring the traffic volume and data exchange rate of the network device according to the inspection cycle, and generating a state index according to the traffic volume and the data exchange rate; If the state index generated during the peak inspection period ∈ the peak regular index interval, then there is no abnormality in the network device, otherwise, there is an abnormality; If the state index generated during the trough inspection period ∈ the trough regular index interval, there is no abnormality in the network device, otherwise, there is an abnormality.

3. The artificial intelligence-based network equipment diagnosis and control method according to claim 2, characterized in that: The process of obtaining the three-dimensional conventional thermal model and the three-dimensional real-time thermal model of the network device includes: Step a1: setting a sampling period; setting a number of infrared thermal imagers around the network device, obtaining infrared thermal imaging videos of the network device in a normal state through the infrared thermal imagers according to the sampling period, and marking the infrared thermal imaging videos as regular thermal imaging videos; Step a2: dividing the conventional thermal imaging video into a number of images in units of frames, marking the images as conventional thermal images, and then generating a conventional thermal image set from conventional thermal images generated by different infrared thermal imagers at the same time; Step a3: Establish a three-dimensional solid model of the network equipment; Step a4: attaching conventional thermal images in the conventional thermal image collection to the three-dimensional entity model of the network device in chronological order to generate a plurality of three-dimensional thermal sensing sub-models; that is, each of the plurality of three-dimensional thermal sensing sub-models has a corresponding time, which is the time corresponding to the conventional thermal image of the three-dimensional thermal sensing sub-model; Step a5: Establish a four-dimensional coordinate system, which includes an X-axis, a Y-axis, a Z-axis and a time axis, and map a number of three-dimensional thermal sensing sub-models to the four-dimensional coordinate system in time sequence, and generate a three-dimensional conventional thermal sensing model through a three-dimensional convolutional neural network; According to the principle of obtaining the three-dimensional normal thermal model of the network device in steps a1 to a5, the three-dimensional real-time thermal model of the network device is obtained.

4. The artificial intelligence-based network equipment diagnosis and control method according to claim 3 is characterized in that: The process of checking whether there are abnormal parts in the 3D real-time thermal model through the 3D conventional thermal model includes: Set the normal feeling interval; Obtain the coordinates corresponding to the RGB value ∈ the common thermal interval in the three-dimensional conventional thermal model, and generate the common thermal area according to the coordinates; Acquire the maximum and minimum values ​​of the normal sensing area on the X-axis, Y-axis and Z-axis according to the four-dimensional coordinate system, and generate the coordinates of the sensing center of the normal sensing area according to the maximum and minimum values; With the centroid coordinate as the center, the X-axis, Y-axis and Z-axis as the value direction, the boundary value change set of the common sensing area is obtained according to the time axis; According to the boundary value variation set, obtain the X value variation range, Y value variation range, Z value variation range, X value variation rate, Y value variation rate and Z value variation rate; Generate a boundary change feature set according to the X value change domain, the Y value change domain, the Z value change domain, the X value change rate, the Y value change rate and the Z value change rate; Obtain the coordinates corresponding to the RGB value ∈ the normal sensing interval in the three-dimensional real-time thermal sensing model, and generate a real sensing area according to the coordinates; The centroid coordinates in the three-dimensional conventional thermal sensing model are obtained, and it is checked whether the centroid coordinates are all within the real sensing area and whether there are centroid coordinates in the real sensing area. If the centroid coordinates are all within the real sensing area and there are centroid coordinates in the real sensing area, a secondary module verification is performed to determine whether there are abnormal parts in the three-dimensional real-time thermal sensing model through the secondary module verification; otherwise, there are abnormal parts. If there are centroid coordinates that are not within the real sensing area, abnormal position information is generated according to the centroid coordinates; if there are centroid coordinates that do not exist in the real sensing area, abnormal position information is generated according to the real sensing area.

5. The artificial intelligence-based network equipment diagnosis and control method according to claim 4, characterized in that: The process of performing secondary module verification to determine whether there are abnormal parts in the 3D real-time thermal model includes: According to the principle process of obtaining the boundary change feature set corresponding to the common sense area corresponding to the sense center coordinate, the boundary change feature set corresponding to the real sense area corresponding to the sense center coordinate is obtained; If the element ranges in the boundary change feature set corresponding to the real sensing area corresponding to the centroid coordinates are all within the corresponding element ranges in the boundary change feature set corresponding to the normal sensing area corresponding to the centroid coordinates, it is determined that there is no abnormal part in the three-dimensional real-time thermal sensing model; otherwise, it is determined that there is an abnormal part in the three-dimensional real-time thermal sensing model; The centroid coordinates corresponding to the abnormal part of the three-dimensional real-time thermal sensing model are obtained, and the abnormal position information is generated according to the centroid coordinates.

6. The artificial intelligence-based network equipment diagnosis and control method according to claim 5, characterized in that: The process of generating corresponding offline fault information and taking corresponding measures according to the offline fault information includes: Offline fault information is generated based on the abnormal location information and sent to the user, informing the user of the suspected abnormal location information so that the user can handle it accordingly.

7. The artificial intelligence-based network equipment diagnosis and control method according to claim 6, characterized in that: The process of taking corresponding measures based on early warning information includes: The warning information is sent to the user to inform the user that there is no offline fault and there is a suspected online fault, and the user takes corresponding measures.

Images