Journal of Thermal Science and Technology (May 2017)
Diagnosis of computer cooling performance based on multipoint temperature measurements
Abstract
We propose a method for inferring the condition of factors influencing the temperature of heat-generating components in electronic equipment. Here, “factor” includes cooling components (fans, heat sinks, etc.), intake temperature, and the caloric value of the heat-generating component. The performance of these cooling components is reduced by unusual use or deterioration due to aging such as clogging or evaporation of grease. Such factors raise the temperature of heat-generating components. High temperatures of heat-generating components lead to system failure or a shortened life of the electronic equipment. Detecting decreases in the performance of cooling components allows timely provision of the required maintenance, possibly leading to increased product reliability. However, monitoring temperature alone is insufficient. The temperature of heat-generating components is also affected by factors such as intake temperature, and caloric value. In this study, we develop a method for inferring the condition of each factor influencing temperatures by using sensors installed at multiple locations. In this method, the effect of each factor on temperature depends on location, and the condition of each factor is expressed as a multivariate function using temperatures at multiple locations. The condition of each factor can then be inferred by measuring only temperatures. Furthermore, we can develop the method simply by mounting several temperature sensors on the board, making the method easy to implement. System failures can thus be prevented by performing maintenance based on the monitored factors. We conducted verification experiments using a computer to confirm the benefits of this method. In the case of inferring the conditions of four factors using four temperature sensors, the mean accuracy exceeded 90%.
Keywords
