The aim of this paper is to create the model for prediction of carbon monoxide release rate (CORR) and smoke production rate (SPR) from heat release rate (HRR) of fast-growing wood species. The model is independent on wood species, thus is suitable for all fast-growing wood species. Three wood species hybrid poplar J-105 (Populus nigra × P. maximowiczii A. Henry), white willow (Salix alba L.) and black locust (Robinia pseudoacacia L.) were used for universal model creation. The heat release rate, smoke production rate and carbon monoxide release rate have been measured at three heat fluxes (25, 35 and 50 kW.m-2) by the cone calorimeter. The average values of CORR and SPR for all investigated wood species were 0.051 g.m-2.s-1 and 0.086 m2.m-2.s-1, respectively. Both dependencies of SPR and CORR on HRR have shown similar trends during the ignition phase (unstable trend) and during the intense burning phase (roughly linear increasing with HRR). The main difference was shown during the steady state phase (dependency of SPR on HRR is stable while dependency of CO on HRR is highly unstable). The results also proved a significant impact of wood density on these dependencies, thus, the neural network for prediction of SPR, CORR from HRR was applied. The coefficients of determination R2 for trained neural networks, for both SPR and CORR, were achieved in the range from 0.96 to 0.97
Read full abstract