A two‐layer temporal stochastic configuration broad learning system for complex industrial soft sensing

Soft Sensor Model Derived from Wiener Model Structure: Modeling and Identification

Unsupervised learning of Self-Localization with Slow Feature Analysis (SFA) using omnidirectional camera input has been shown to be a viable alternative to established SLAM approaches. Previous models for SFA self-localization purely relied on omnidirectional visual input. The model led to globally consistent localization in SFA space but the lack of odometry integration reduced the local accuracy. However, odometry integration and other downstream usage of localization require a common coordinate system, which previously was based on an external metric ground truth measurement system. Here, we show an autonomous unsupervised approach to generate accurate metric representations from SFA outputs without external sensors. We assume locally linear trajectories of a robot, which is consistent with, for example, driving patterns of robotic lawn mowers. This geometric constraint allows a formulation of an optimization problem for the regression from slow feature values to the robot’s position. We show that the resulting accuracy on test data is comparable to supervised regression based on external sensors. Based on this result, using a Kalman filter for fusion of SFA localization and odometry is shown to further increase localization accuracy over the supervised regression model.

In the field of freeway traffic safety research, there is an increasing focus in studies on how to reduce the frequency and severity of traffic crashes. Although many studies divide factors into “human-vehicle-road-environment” and other dimensions to construct models whichshowthe characteristic patterns of each factor's influence on crash severity, there is still a lack of research on the interaction effect of road and environment characteristics on the severity of a freeway traffic crash. This research aims to explore the influence of road and environmental factors on the severity of a freeway traffic crash and establish a prediction model towards freeway traffic crash severity. Firstly, the obtained historical traffic crash data variables were screened, and 11 influencing factors were summarized from the perspective of road and environment, and the related variables were discretized. Furthermore, the XGBoost (eXtreme Gradient Boosting) model was established, and the SHAP (SHapley Additive exPlanation) value was introduced to interpret the XGBoost model; the importance ranking of the influence degree of each feature towards the target variables and the visualization of the global influence of each feature towards the target variables were both obtained. Then, the Bayesian network-based freeway traffic crash severity prediction model was constructed via the selected variables and their values, and the learning and prediction accuracy of the model were verified. Finally, based on the data of the case study, the prediction model was applied to predict the crash severity considering the interaction effect of various factors in road and environment dimensions. The results show that the characteristic variables of road side protection facility type (RSP), road section type (LAN), central isolation facility (CIF), lighting condition (LIG), and crash occurrence time (TIM) have significant effects on the traffic crash prediction model; the prediction performance of the model considering the interaction of road and environment is better than that of the model considering the influence of single condition; the prediction accuracy of XGBoost-Bayesian Network Model proposed in this research can reach 89.05%. The identification and prediction of traffic crash risk is a prerequisite for safety improvement, and the model proposed and results obtained in this research can provide a theoretical basis for related departments in freeway safety management.

Learning representations that disentangle hidden explanatory factors in data has proven beneficial for effective pattern classification. Slow feature analysis (SFA) is a nonlinear dimensionality reduction technique that provides a useful representation for classification if the training data is sequential and transitions between classes are rare. The pattern discrimination ability of SFA has been attributed to the equivalence of linear SFA and linear discriminant analysis (LDA) under certain conditions. LDA, however, is often outperformed by locality preserving projections (LPP) when the data lies on or near a low-dimensional manifold. Here, we take a unified manifold learning perspective on LPP, LDA and SFA. We suggest that the discrimination ability of SFA is better explained by its relation to LPP than to LDA, and give an example of a situation where linear SFA outperforms LDA. We then propose a novel supervised manifold learning architecture that combines hierarchical nonlinear expansions, as commonly used for SFA, with supervised LPP. It learns a nonlinear parametric data representation that explicitly takes both the class labels and the manifold structure of the data into account. As an experimental validation, we show that this approach outperforms previously proposed models on the NORB object recognition dataset.

Accurate monitoring of complex industrial plants is crucial for ensuring safe operations and reliable management of desired quality. Early detection of abnormal events is essential to preempt serious consequences, enhance system performance, and reduce manufacturing costs. In this work, we propose a novel methodology for fault detection based on Slow Feature Analysis (SFA) tailored for time series models and statistical process control. Fault detection is critical in process monitoring and can ensure that systems operate efficiently and safely. This study investigates the effectiveness of various multivariate statistical methods, including Slow Feature Analysis (SFA), Kernel Slow Feature Analysis (KSFA), Dynamic Slow Feature Analysis (DSFA), and Principal Component Analysis (PCA) in detecting faults within the Tennessee Eastman (TE), Benchmark Simulation Model No. 1 (BSM 1) datasets and Beijing wastewater treatment plant (real world). Our comprehensive analysis indicates that KSFA and DSFA significantly outperform traditional methods by providing enhanced sensitivity and fault detection capabilities, particularly in complex, nonlinear, and dynamic data environments. The comparative analysis underscores the superior performance of KSFA and DSFA in capturing comprehensive process behavior, making them robust, cutting-edge choices for advanced fault detection applications. Such methodologies promise substantial improvements in industrial plant monitoring, contributing to heightened system reliability, safety, and overall operational efficiency.

Exposing and understanding synergistic effects in co-pyrolysis of biomass and plastic waste via machine learning

Data‐driven soft sensor models have been extensively utilized in industrial processes. Batch processes are usually employed to manufacture low‐volume and high value‐added products in chemical, materials, and pharmaceutical industries. The most distinctive features of batch process lie in nonlinear, repetition, and slow time varying characteristics. In this paper, a data‐driven soft sensor modelling method based on linear slow feature analysis (LSFA) and least squares support vector regression (LSSVR) is proposed. In this method, LSFA was used to effectively capture the driving force behind the data features that change as slowly as possible. Then, a LSSVR model was constructed between the extracted slow feature variables and quality variables. Finally, a numerical example, industrial penicillin fermentation processes, and cobalt oxalate synthesis process were utilized to confirm the prediction accuracy and model reliability of the proposed approach.

Long-term dependency slow feature analysis for dynamic process monitoring

Slow feature analysis-independent component analysis based integrated monitoring approach for industrial processes incorporating dynamic and static characteristics

Slow feature analysis (SFA) can extract the features representing intrinsic properties. Moreover, in an actual industrial process: information can be found in low-frequency data signals. Hence, SFA detects subtle changes more sensitively than other algorithms do. On this basis, a dynamic total slow feature regression is proposed to achieve efficient quality-relevant fault detection for dynamic processes. Initially, lagged measurements are introduced to process variables as additional variables for further obtaining the dynamic information of the industrial process. Then, to solve the information redundancy problem and improve the quality-relevant fault detection performance, total projection method is utilized to divide the process variables into two parts, which are relevant to and independent from quality. Finally, monitoring statistics on two subspaces are constructed to detect quality-relevant and irrelevant faults. The experiment results on the Tennessee Eastman process demonstrates that the proposed method outperforms some other state-of-the-art methods.

Hybrid change detection (HCD) for high-resolution imagery usually adopt decision-level method and rely on artificial design. To address this issue, we propose a novel feature-level fusion strategy for HCD based on iterative slow feature analysis (ISFA). First, objects are obtained by multiresolution segmentation of bi-temporal images respectively, and corresponding feature sets are constructed through stacking pixel- and object-level spectral features. Then, slow feature analysis (SFA) is used for transforming the feature sets into a new feature space at the first time. And iteration method with variable weights is introduced to get the last slow feature fusion map, where the changed pixels and unchanged pixels can be separated more easily. At last, K-means cluster is adopted to separate changed area and unchanged area automatically and generate final change result. Experiments were conducted on bi-temporal multi-spectral images, demonstrating the good performance of the proposed approach.

In this paper we propose a novel face recognition approach based on slow feature analysis (SFA) in contourlet transform domain. This method firstly use contourlet transform to decompose the face image into low frequency and high frequency part, and then takes technological advantages of slow feature analysis for facial feature extraction. We named the new method combining the slow feature analysis and contourlet transform as CT-SFA. The experimental results on international standard face database demonstrate that the new face recognition method is effective and competitive.

As a key component of the underwater operation platform, the accurate prediction of the remaining useful life of the underwater platform can not only reduce the maintenance cost of the equipment, but also improve the intelligence of equipment management. In this paper, the remaining useful life prediction method based on SSA-SVR is proposed, on this basis, the slow feature analysis method is combined to improve the prediction accuracy of the remaining useful life of the underwater turntable. Finally, the prediction results of SSA-SVR based on slow feature analysis are compared with those of SSA-SVR based on wavelet transform. The results show that the remaining useful life prediction method based on slow feature analysis is better than that of wavelet transform.

Extracting invariant features in an unsupervised manner is crucial to perform complex computation such as object recognition, analyzing music or understanding speech. While various algorithms have been proposed to perform such a task, Slow Feature Analysis (SFA) uses time as a means of detecting those invariants, extracting the slowly time-varying components in the input signals. In this work, we address the question of how such an algorithm can be implemented by neurons, and apply it in the context of audio stimuli. We propose a projected gradient implementation of SFA that can be adapted to a Hebbian like learning rule dealing with biologically plausible neuron models. Furthermore, we show that a Spike-Timing Dependent Plasticity learning rule, shaped as a smoothed second derivative, implements SFA for spiking neurons. The theory is supported by numerical simulations, and to illustrate a simple use of SFA, we have applied it to auditory signals. We show that a single SFA neuron can learn to extract the tempo in sound recordings.

The past decade has seen a rise of interest in Laplacian eigenmaps (LEMs) for nonlinear dimensionality reduction. LEMs have been used in spectral clustering, in semisupervised learning, and for providing efficient state representations for reinforcement learning. Here, we show that LEMs are closely related to slow feature analysis (SFA), a biologically inspired, unsupervised learning algorithm originally designed for learning invariant visual representations. We show that SFA can be interpreted as a function approximation of LEMs, where the topological neighborhoods required for LEMs are implicitly defined by the temporal structure of the data. Based on this relation, we propose a generalization of SFA to arbitrary neighborhood relations and demonstrate its applicability for spectral clustering. Finally, we review previous work with the goal of providing a unifying view on SFA and LEMs.