Guest Editorial: Special issue on computational intelligence for industrial data processing and analysis
暂无分享,去创建一个
Data processing and analysis play a key role in process industries, where nonlinear dynamics with various uncertainties present and mathematical model-based approaches fail. Computational intelligence (CI) techniques, including fuzzy logic systems, artificial neural networks and evolutionary computing, have demonstrated great power to deal with data modeling through sample learning, knowledge representation and structure/parameter optimization. Over the past decades, research on CI-based intelligent systems has received considerable attention and many successful real-world applications have been reported in the literature. For instance, in control engineering, unknown nonlinear dynamics can be modeled by using neural networks and different control schemes can be developed based on learner models; fuzzy logic controllers can be implemented by using adaptive neuro-fuzzy inference systems; and fault diagnosis and prognosis to improve the operational safety, reliability and maintainability of engineering systems in critical conditions. With the increasing complexity of modern engineering systems, we are still facing many challenges in processing industrial data, such as the extraction of features from a large amount of measurements with a distributed nature, learning algorithms for robust data modeling, signal processing-based diagnosis system design, and the integration of system modeling, control and diagnosis. The papers published in this special issue are a collection of selective submissions from the 11th World Congress on Intelligent Control and Automation (WCICA2014), June 27–30, 2014, Shenyang, China. This special issue focuses on the recent promising advancement of industrial applications of computational intelligence systems. After extensive reviews and revisions, 11 papers have been accepted. These papers cover the most active areas of research and applications in industrial data processing and analysis, including a randomized algorithm for robust data modeling for the mineral industry, image processing systems, multipleobjective operation optimization for electric multiple units, highdimensional feature reduction for temporospatial data, fault diagnosis and wind prediction of power systems. We tried our best to make this special issue informative and high in quality and significance, as such a collection will be useful and valuable to both researchers and engineers. We provide a summary of the papers published in this special issue as below. The paper entitled “Particle size estimate of grinding processes using random vector functional link networks with improved robustness”, by W. Dai, Q. Liu and T. Y. Chai, presents a robust modeling approach for the prediction of an important product quality index in grinding processes, where the particle size (PS) is usually established by using the population balance method (PBM). However, the parameters of the PBM-based model are more empirical in nature and even unknown, whereas the datadriven models are often unsatisfactory due to the uncertainties in choosing an appropriate model structure and parameters and obtaining sufficient training data to represent all the process behavior. To address these problems, this paper proposes a hybrid PS model, which is composed of a mechanism model and a random vector functional link network (RVFLN)-based compensation model. Due to the fact that the model quality of traditional RVFLN may deteriorate whenever the training data is contaminated with outliers, a robust learning algorithm for building RVFLNs is proposed to improve the modeling performance. The robust RVFLN incorporates a kernel density estimation-based weighted least squares method into the loss function. The recursive extension is also presented to reduce the memory space and computational load of the traditional RVFLN learning techniques. The application results on a hardware-in-the-loop experiment system of grinding processes demonstrate the effectiveness and the improved robustness of the proposed modeling techniques. The paper entitled “2-D regularized locality preserving projection algorithms for temporospatial feature reduction and its application in industrial data regression”, by X. J. Zhou, D. H. Wang and Z. G. Shao, develops dynamic soft sensors for measuring product quality in complex industrial processes. The input variables of the proposed regression model are composed of temporospatial data with some redundancies, which may result in a poor generalization performance of the soft sensor. This paper proposes a two-dimensional regularized locality preserving projection (2DRLPP) algorithm for feature reduction, which combines the locality preserving projection (LPP) method with data roughness regularization. An extension of the proposed 2DRLPP is given and termed bidirectional 2DRLPP. A case study on