Purpose The purpose of this paper is to improve the dyeing effect of fast-growing fir wood dyed with reactive dyes. Design/methodology/approach In this study, five factors including temperature, the dosage of dye accelerator, dyeing time, the dosage of fixing agent and fixing time were investigated. Then, the color difference and light resistance of the wood surface after dyeing were used as the evaluation indicators; the best dyeing process under the two indicators was obtained through the range analysis. Finally, the two indicators were considered comprehensively, and the fuzzy comprehensive evaluation method was used to obtain the best dyeing process under the comprehensive indicators. Findings The results show that when the comprehensive index was used as the evaluation index, the optimal dyeing process for reactive red X-3B dyeing fast-growing fir veneer was that the dyeing temperature was 65°C; the amount of dye accelerator was 25 g L−1; the dyeing time was 2 h; the amount of fixing agent was 15 g L−1; and the fixing time was 35 min. Originality/value The technique of wood dyeing is an important method to increase the value of wood products. When using different kinds of dyes or dyeing substrates for wood dyeing, the dyeing process is different. This study determined the best process for reactive dye dyeing of fast-growing fir veneer and provided a solution for improving the value of fast-growing fir wood.

Wood dyeing technology is of great significance to improve the utilization rate of inferior wood resources. The challenge to imitating precious wood species by inferior wood is to quickly and accurately obtain the dyeing formula of precious wood species. This study uses Genetic Algorithm (GA) to optimize Extreme learning machine (ELM), and then a predictive model based on GA-ELM is proposed for predicting the dyeing formula of precious wood species. The sum of the relative deviations of the three dyes between the predicted formula and the actual formula, that is, the relative deviation of the formula, is calculated to evaluate the model’s prediction accuracy. The simulation results show that the average relative deviations of the formula predicted by Back Propagation (BP) neural network, Radial Basis Function (RBF) neural network, ELM, and GA-ELM are 0.808, 0.717, 0.708, and 0.262. The prediction deviation of the GA-ELM is much smaller than that of other traditional neural networks, which can achieve good results in wood production.