Fine-grained Temperature Monitoring System For Server Room Equipment Based On Infrared Thermal Imaging

Nowadays,server rooms generally concentrate a large amount of electronic equip-ment and computing resources,resulting in significant heat generation.However,due to the negligible impact of the temperature anomaly of a single device on the overall ambient temperature of the server room,it can be difficult to detect this in a timely manner,which can easily lead to irreparable losses.Therefore,this paper designs a fine-grained tem-perature monitoring system for server room equipment.This system combines infrared thermal imaging technology and a rule base for judging whether a single device is ab-normal to conduct flexible and adaptable monitoring and early warning for each piece of equipment in the server room.The paper employs engineering application technologies and supplements them with time series forecasting algorithms to design and implement a fine-grained temperature monitoring system for server room equipment.The main contributions include:1)Accurate real-time measurement results.By combining infrared cameras with other hardware devices,real-time monitoring of equipment in the computer room is achieved to achieve fine-grained temperature monitoring,ensuring that every temperature of the equipment can be monitored.2)Flexible and adaptable monitoring effects.The system has established an open ab-normal temperature rule library in terms of functionality,initially setting hard rules,and incorporating historical data into the time dimension to form time series.Rel-evant algorithms are used for training to form soft rules as a supplement to hard rules.Ultimately,each device can choose to create different monitoring schemes according to their needs.3)Intelligent monitoring and early warning functions.By introducing time series pre-diction algorithm models such as ARIMA,CNN,LSTM,Transformer,and CNN-Bi LSTM-Attention,it learns from temperature data sets at different sampling fre-quencies on specific devices,thereby accurately defining abnormal temperatures,and realizing adaptive temperature detection and intelligent early warnings for indi-vidual devices.Experimental results show that under mean square error,root mean square error,mean absolute error,and mean absolute percentage error indicators,the optimal models for sampling frequencies of 1 minute,5 minutes,10 minutes,30 minutes,60 minutes are LSTM,ARIMA,and CNN respectively,keeping the temperature forecast error within 0.5℃.Finally,the system’s functionality and performance are tested,verifying the usability of system functions and the robustness of performance.