A method for monitoring and diagnosing server thermal faults based on infrared images

Abstract

The invention discloses a method for monitoring and diagnosing server thermal faults based on infrared images, which comprises: collecting infrared images and visible light images of servers under different thermal fault states; Perform tilt correction processing on the image below, and use the manual segmentation method to determine the region of interest containing the server in the infrared image; convert the determined region of interest image into a corresponding grayscale image, and extract features based on image entropy; The PCA method performs dimensionality reduction on the obtained global entropy and local entropy row-column mean features; the feature principal components obtained after dimensionality reduction are used to train the support vector machine classifier; the infrared images and visible light images corresponding to the servers to be detected are collected, The thermal fault state type of the server is diagnosed by the trained SVM model. The invention greatly improves the management efficiency of the data center management personnel, and facilitates engineering debugging and system maintenance.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method for monitoring and diagnosing thermal faults of servers based on infrared images. Background technique [0002] While the scale of the data center is expanding rapidly, the high-density integration of the server leads to an increase in its heat generation. In order to make the server run normally in a suitable environment without causing downtime accidents caused by excessive temperature, many data centers have to increase cooling. The power of the air conditioner causes high energy consumption and low efficiency in the data center. Therefore, accurately monitoring and controlling the temperature of the computer room, improving the cooling efficiency of the air conditioner, and realizing the energy saving and consumption reduction of the data center have become issues that cannot be ignored in the process of data center construction and maintenance. ...

AI Technical Summary

Problems solved by technology

At the same time, the spatial resolution of the temperature sensor is insufficient to obtain detailed information on hot spots

If the sensor is not deployed in a suitable position, we cannot accurately obtain the size and maximum temperature of the hot spot, which will not provide sufficient information for subsequent processing

Images