Neuromorphic Random Walk for Experimental Phosphate Adsorption Modeling

Neuromorphic systems leverage their intrinsic parallelism and event-driven behavior to deliver energy-efficient and low-latency solutions. Although such advantages have been extensively demonstrated in proof-of-concept settings with synthetic data, evidence of high performance on real-world tasks remains scarce. To address this gap, we apply an unsupervised neuromorphic random walk (NRW) algorithm to experimental data from phosphate adsorption on multiple substrates. Our model autonomously segments sensor dynamics into regimes that agree with the underlying adsorption theory, with a strong fit ($R^2 > 0.98$) with the traditional pseudo-second-order (PSO) kinetics model used in this case. Using the Langmuir model, the NRW also predicts consistent Gibbs free energy values for this process. This work provides an initial step towards a theory-consistent and practical NRW deployment, enabling more efficient and noise-resilient sensing platforms.