A Novel Machine Learning-Based Quantification of Frost Targeted for Energy Conservation in Refrigeration Systems

Abstract

Energy conservation is required in refrigeration systems. The frost formation on the fin tube evaporator of the common refrigeration system acts as an insulator, reducing the heat transfer efficiency. Excessive ice or frost can block the airflow across the evaporator coils, reducing the air circulation within the refrigerator compartment. Ice or frost formation on the evaporator coils leads to reduced cooling within the refrigerator compartment, affecting the freshness and quality of stored items. Continuous ice or frost accumulation on the evaporator coils can also cause long-term damage to the fin tubes. This can result in reduced efficiency and eventually lead to the failure of the evaporator. To address these issues, refrigerators employ various active and passive defrosting techniques such as time-based defrost cycles and temperature-based defrost cycles. These methods initiate defrosting to remove the ice or frost from the evaporator coils. At times even when there is no frost, the defrosting techniques are active which results in power loss, hence the main problem is to detect whether sufficient frost is present or not and based on the result defrosting can be applied. I suggest employing a machine learning based approach on the Convolutional Neural Networks (CNN) rather than the conventional frost detection approaches to accurately quantify frost so that defrosting can be applied at accurate time interval, which will result in conservation of energy. A Refrigeration engineer may find it easier to distinguish frost using deep learning algorithms that have been trained on data set and help them learn and extract meaningful representations from the input data through CNN. These representations enable the network to make accurate predictions or classifications-based features, ensuring efficient cooling and optimal performance of the refrigerator.