Developing an experiment-based strong machine learning model for performance prediction and full analysis of Maisotsenko dewpoint evaporative air cooler

Among all direct and indirect water-based evaporative air coolers, Maisotsenko indirect evaporative cooler (MIEC) is the only one which can supply sub wet-bulb air temperature while adding no moisture to the product air. For any given MIEC, four thermal/fluid parameters including dry channel flow rate, wet-to-dry-side air flow ratio, ambient air temperature and ambient relative humidity impact the performance of the MIEC. Sub-wet bulb supply temperature while having maximum cooling capacity happen only for a specific range of the mentioned thermal/fluid parameters. For any ambient condition and dry side air flow, a magic value of air-flow-ratio can maximize the cooling capacity of the cooler. Thousands of data is required to provide a simultaneous multi-aspect analysis of each parameter (while other parameters vary as well) to identify all possible curve behaviors and secrets of this technology. That is why, for the first time in this research, a strong machine learning (ML) model, is developed for MIEC cooler using 625 validated analytical/experimental-based training data (full factorial design). This model is employed to generate around 3600 unseen data to fully clarify the impact of each parameter (on supply temperature, total cooling capacity, useful cooling capacity and wet-bulb effectiveness) while other parameters are variable too. This strategy makes it possible to see all hidden extremum or peak point where the performance of the cooler is maximized or minimized. Interesting results are found in this research. For any ambient condition, there is an extremum point of the air flow ratio at which the total cooling capacity is maximized. The results indicate that as the dry side air flow increases, this extremum value decreases. For instance, under an ambient condition of 18 °C and 9 % relative humidity, the extremum point of the air flow ratio is 0.45 when the dry side air flow is 0.8 L/s and 0.25 when the dry side air flow is 4.8 L/s. In real-world commercial cooler, this can be simply happened by appropriate re-adjusting the dry and wet side fan speeds. • Maisotsenko evaporative cooler can supply sub-wet-bulb temperature. • Machine Learning model is developed for this technology. • All possible behaviours of this cooler were recognized. • The cooling capacity is maximized under specific range of fluid-flow factors. • Magic number of air flow ratio is identified, analysed and reported.