Abstract

A natural rubber sponge (NRP) was prepared by varying the carbon black, blowing agent, and sulfur contents and subsequently used as a precursor for biocarbon sponge (BC) production. The experiments and the machine learning approach used an adaptive multi-input, fuzzy-rules, emulated network (MiRREN) to optimize the NRP formulation to achieve the best BC properties. Based on the results, adding carbon black stabilized the shape of the BC samples after carbonization. The BC samples had an interconnected pore structure and oxygen and sulfur functionalities. The standard isotherm of the sample without carbon black was type IV, which was transformed to type II after adding carbon black, with a wide range in the pore size distribution. The addition of carbon black improved the electrical conductivity from 1.13 S cm⁻¹ to a maximum value of 12.78 S cm⁻¹. At 0.25 A g⁻¹, the samples’ specific capacitance was enhanced to 152.7 and 137.9 F g⁻¹ using certain conditions with carbon black. The sample with 50, 6, and 20 phr of carbon black, blowing agent, and sulfur contents, respectively, had the best cyclic performance (113.7 %) at 10,000 cycles of testing, verifying the excellent cyclic durability due to the surface functionalities and the addition of the carbon black, as well as the developed interconnected pore structure. Based on the results of the machine learning study, the pore volume and specific capacitance were likely correlated with the blowing agent content rather than the sulfur content. The optimum formulation of NRP for achieving the best specific capacitance levels at all the applied current densities was 7.02 phr of blowing agent and 20.2 phr of sulfur content.