As a pivotal role in the control, optimization, and monitoring of contemporary industrial processes, soft sensors are frequently employed in the prediction of key quality variables. To achieve accurate prediction of key quality variables in industrial processes, a soft sensor modeling method based on the self-organizing fuzzy neural network with the clustering, merging, and splitting scheme (SOFNN-CMS) is proposed. First, the supervised fuzzy C-means clustering algorithm is proposed to identify the appropriate initial center and width of the fuzzy neural network, obtaining appropriate initial fuzzy rules. Then, a neuron merging and splitting strategy is designed to adjust the structure of the fuzzy neural network, by merging and splitting the hidden neurons according to the distance of clusters, increasing the adaptability of the fuzzy neural network. Besides, to accelerate the convergence of estimation errors, an improved Levenberg Marquardt algorithm is utilized to update neural network parameters in the training phase, realizing the soft sensor modeling of key quality variables. The effectiveness of the proposed SOFNN-CMS neural network is demonstrated on two benchmark problems and an industrial debutanizer column. Finally, the experiments showcase that the proposed SOFNN-CMS neural network can obtain better soft sensor modeling performance with a compact structure.
11 August 2025 in IEEE Transactions on Fuzzy Systems
To tackle the insufficient representation of uncertainty and weak structural adaptability in nonlinear system modeling, a dynamic interval type-2 fuzzy stochastic configuration network (DIT2FSCN) is proposed, synergistically integrating fuzzy rule initialization and dynamic sparse interpolation rule growth. First, a k-means clustering algorithm based on density peaks (BDPKA) is introduced to initialize the parameters of DIT2FSCN. The limitations of traditional clustering methods, such as sensitivity to initial centers and the requirement to predefine the number of rules, can be overcome by fusing density peak detection with k-means clustering, enabling data-driven fuzzy rule generation. Second, a two-stage modeling framework incorporating interval type-2 fuzzy inference and stochastic configuration-based dynamic optimization is designed. The uncertainty modeling capability of an interval type-2 fuzzy logic system (IT2FLS) and the incremental learning ability of stochastic configuration networks are jointly harnessed, allowing for dynamic characterization of complex system behaviors. Third, a dynamic sparse interpolation rule growth strategy based on the blankness score is proposed. The blankness scoring mechanism that accounts for both inter-rule spacing and local data distribution is designed to facilitate adaptive rule bases expansion via sparse region identification and dynamic interpolation, capturing the coupling relationships in nonlinear systems. Finally, the effectiveness of DIT2FSCN is validated through simulation experiments, including sequence prediction, benchmark dataset modeling, and two real-world industrial process applications. The experimental results demonstrate that the proposed DIT2FSCN has better performance than its peers in accuracy and generalization capability.