Microalgal density modeling using a low-cost spectral sensor: A Kolmogorov–Arnold network approach

Microalgae cultivation has recently captured several applications in the field of biotechnology. Microalgae produce biomass that can be converted into a variety of valuable products, including biofuels, bioplastics, dietary supplements, and feed for aquaculture. In microalgae cultivation, monitoring microalgal density estimation is necessary. Traditional microalgae estimation techniques, such as manual counting and the use of a spectrophotometer, are prone to error, costly, and time-consuming. In this paper, we develop a low-cost integrated circuit (IC)-based spectral sensor with Kolmogorov–Arnold network (KAN) as the estimator model for density. The results indicate that the proposed KAN outperforms feedforward neural networks (FNNs). It shows that, despite its limited resolution, the sensor is capable of achieving acceptable performance. • A low-cost sensor for estimating microalgae density is design. • Kolmogorov–Arnold network is employed as the model estimator. • The design system achieves the best RMSE of about 0.5 for both Chlorella vulgaris and Monoraphidium contortum . • The design system achieves the best R 2 of 0.9996 for both Chlorella vulgaris and Monoraphidium contortum .