The complex chemical process has problems of strong nonlinear, multivariable coupling, parameters time-varying and large time delay, which result in low accuracy of soft sensor. To address these issues, a soft sensor modeling method based on the regularization AdaBound interval type-2 fuzzy neural network (RAIT2FNN) is proposed. Firstly, to solve the problem that the structure of interval type-2 neural network ( IT2FNN) is difficult to determine, an algorithm for self-organizing generation rules that uses firing strength and rule similarity to define growth and deletion indicators is proposed. The algorithm uses the firing strength to determine whether to generate rules, and deletes the rules according to the similarity. In this way, the architecture of the IT2FNN is determined. Secondly, this article proposes AdaBound with regularization to modify the relevant parameters of the RAIT2FNN model. And different parameters have bounded adaptive learning rates. Finally, RAIT2FNN is used as a soft sensor model to predict the tail oxygen concentration for uncatalysed oxidation of cyclohexane process. The experimental results are that the test time is 0. 008 2, the training RMSE is 0. 018 2, and the test RMSE is 0. 009 6, indicating that RAIT2FNN as a soft sensor model has the advantages of timely prediction and high prediction accuracy.