1. INTRODUCTION
Microalgal biodiesel promises to be a zero CO2-emitting and renewable energy source. The biodiesel production from microalgae includes various steps beginning from cultivation of the desired strain followed by harvesting, drying of microalgal slurry, oil extraction from the biomass, and transesterification of lipid to biodiesel (O’Connell, Savelski, & Slater, 2013). However, the downstream processing of the harvested microalgal slurry faces several technical challenges for cost-effective biodiesel production. Drying of microalgal biomass is one of those important steps. Dewatering of the harvested microalgal slurry is essential to enhance the viability of biomass for lipid extraction, and also to reduce the costs of handling, transportation, packaging, and storage (Bennamoun, Arlabosse, & Léonard, 2013).
Many drying techniques are available nowadays which include advanced techniques like spray drying, tray drying, oven drying, freeze drying, vacuum-shelf drying, cross-flow air drying, etc, as well as the simple and natural solar drying. The suitability of a method depends on many factors such as the properties and types of the microalgal suspension, the downstream processes, the end product type and quality, and the overall cost of production (Chen, Chang, & Lee, 2015). The drying techniques such as freeze drying (Guldhe, Singh, Rawat, Ramluckan, & Bux, 2014), thin layer drying (Hosseinizand , Sokhansanj, & Lim, 2018), microwave drying (Villagracia et al., 2016) etc. are mainly used at a laboratory-scale, whereas drum drying technique could be useful at industrial level for drying of large-scale microalgal slurry.
The major advantages of drying the microalgal slurry in a drum dryer are breaking the cell wall and sterilization of the dried biomass (Show, Lee, & Mujumdar, 2015). Mahadevaswamy & Venkataraman (1981) reported that drum drying of Scenedesmus acutus 273-3a biomass for a detention time of 8-10 sec reduced the microbial load during storage. An electrically-heated drum dryer was tested for drying ofScenedesmus sp. biomass containing 70% moisture for 10 sec at 120 °C and the energy consumption was calculated to be 52 kWh (Becker & Venkataraman, 1982). A steam-heated drum drying of Spirulinaslurry was carried out at 120-128 °C for 16 sec for preparing the feed for aquaculture (Sethi & Naik, 2007). Recently, Wahlen et al. (2017), used a bench-scale rotary drum dryer for drying the Scenedesmussp. at 50 0C. Thus, for faster drying, a drum dryer could be a suitable option for large-scale processing of microalgal slurry.
Machine learning, on the otherhand is a new edge technology used in every sector of research. It can be divided into two segments, i.e. supervised learning and unsupervised learning. Supervised learning is implemented when a structured data set is provided and we know what our output would look like. Whereas, unsupervised learning allows us to tackle the problem with little or no knowledge of what our results would be. It is further divided as shown in the Figure 1.
There are different algorithms to implement the mentioned modules (Figure 1). Regression can be done using linear regression algorithm and classification using logistic regression or simple vector machine. Algorithms like neural networks can be used for both. Unsupervised learning can be done through K-mean or principal component analysis.
For our defined problem, we have taken the supervised-classification module into consideration to label the outputs good or poor. Out of logistic regression and simple vector machine, logistic regression is employed based on the fact that we have a large number of features compared to training examples. Here, there was a chance for overfitting of the decision boundary for which regularization was also executed. Logistic regression is a predictive algorithm. The present investigation is thus, aimed at developing a microalgal drying protocol at different drum surface temperature with varying drum speed using the logistic regression with regularization to obtain maximum lipid from microalgal biomass for biodiesel purpose.