Data-driven Modeling Method Research Articles

A Training-Free Data-Driven Method for Input-Output Modeling of Complex Processes

In a variety of human-in-the-loop systems, variations among human operators can result in inconsistencies in process operation and product quality. While a variety of methods exist to mitigate this issue, they often require some model of the relationship between the human input and system output; unfortunately, obtaining such a model continues to be very difficult for highly complex processes such as industrial manufacturing processes. In this paper, we propose an innovative training-free data-driven (TFDD) modeling method that directly predicts the next state from the state transition information of all samples in a database. Because the prediction is directly derived from the database, the model does not require any training, nor does the model architecture change from one application to another. Through a case study on human operator supervisory control of twin-roll steel strip casting, we demonstrate the performance and advantages of the proposed TFDD method as compared to a baseline nonlinear autoregressive network with exogenous inputs (NARX) model trained using the same dataset.

Systematic Comparison of Supervised Learning Methods to Reduce Calibration Effort in Engine Control Development

Complexity of engine control systems is continuously growing due to an increased number of subsystems and the need for robust performance. For traditional map-based as well as state-of-the-art model-based approaches, this will lead to unacceptable development costs and time for future engines. Parametrization of the embedded models using supervised learning regression methods can immensely reduce the number of calibration parameters and hence the calibration effort. However, a methodology for performance assessment of different promising data-driven modelling methods for engine control development is currently lacking. In this paper, a systematic methodology that assesses model inaccuracy, and also implementation aspects such as calibration effort and computational complexity is introduced. This method is applied to assess the potential of Supervised Learning (SL) methods for parametrizing the feedforward controller of a modern diesel engine air-path controller. Using requirements analysis and the specified performance criteria, two regression methods were selected: artificial neural networks (ANN) and support vector machines (SVM). After careful data selection and model training, performance is compared with the benchmark controller, which uses a physics-based model. From simulation results, it is shown that a 97% reduction in the number of calibration parameters with both regression models can be realized. For a standard test cycle, cumulative engine out NOx emissions with regression based controllers are close to the allowable inaccuracy of 10% compared to the benchmark controller. Among the two methods, ANN shows the best performance for the studied performance criteria of inaccuracy, number of calibration parameters and computational complexity.

A Novel Data-Driven Modeling Method for the Spatial–Temporal Correlated Complex Sea Clutter

The sea clutter, referred to as the time-varying radar backscatter from the ocean surface, plays a significant role in marine radar development. The ocean’s complex hydrodynamics cause it to exhibit non-Gaussian and nonstationary characteristics, which brings challenges in the sea clutter modeling, especially for establishing its spatial–temporal correlated and coherent model. In this article, a data-driven method based on the Koopman mode decomposition (KMD) is proposed for modeling spatial–temporal correlated complex sea clutter. The method decomposes the coherent sea clutter dynamic behavior in terms of Koopman modes and corresponding temporal patterns. Then, these spatiotemporal patterns are used to construct the sea clutter state over time according to the approximate solution. Furthermore, this proposed state-of-the-art data-driven approach is benchmarked by the measured sea clutter data from intelligent PIXel processing radar (IPIX). It is demonstrated that the proposed approach accurately models the complex sea clutter with actual statistic characteristics, phase information, and spatial–temporal correlations. The mean absolute error (MAE) and root mean square error (RMSE) between the obtained and actual sea clutter are only 0.1817 and 0.2349, respectively. This work offers a practical approach for modeling sea clutter, especially when the spatial–temporal correlation and coherence information is needed.

Optimal Control Design for Traffic Flow Maximization Based on Data-Driven Modeling Method

This paper proposes enhanced prediction and control design methods for improving traffic flow with human-driven and automated vehicles. To achieve accurate prediction for the entire time horizon, data-driven and model-based prediction methods were integrated. The goal of the integration was to accurately predict the outflow of the traffic network, which was selected as the highway section in this paper. The proposed novel prediction method was used in the optimal design for calculating controlled inflows on highway ramps. The goal of the design was to reach the maximum outflow of the traffic network, even against disturbances on uncontrolled inflows of the network. The control design leads to an optimization problem based on the min–max principle, i.e., the traffic outflow is considered to be maximized by controlled inflows and to be minimized by uncontrolled inflows. The effectiveness of the prediction and the control methods through simulation examples are illustrated, i.e., traffic outflow can be maximized by the control system under various uncontrolled inflow values.

The Use of Corpora in the Teaching of ESP (the Sphere of Pedagogy and Psychology)

Specialized language from the spheres of pedagogy and psychology constitutes a fundamental aspect within the teacher – student communication. The submitted paper handles lexical and morphological features of compound nouns containing the nouns amnesia, memory and recall. The primary research interest focuses on collocations and concordances in which they appear on a regular basis; we also deal with the grammatical relations and elements of meaning that have an impact on the discourse characteristics. The studied lexical units are elaborated within English Web 2015 utilizing Sketch Engine. In the research, both quantitative (statistical methods) and qualitative (observation, comparison, generalization, data-driven research data modelling) methods were employed. To begin with, we provide the frequency list of the studied words within the text corpus, and we categorize them from the morphological perspective as the two aspects influence their lexical behaviour and employment for teaching purposes. The research outcomes indicate that the analysed lexical units exist in a broad discourse scope; moreover, they appreciably grant profitable acts of communication in the pedagogical settings. Eventually, we outline feasible pedagogical inferences of the inquiry outcomes in the teaching of English and we suggest a set of corpus-driven exercises on the professionally-oriented language from the fields of pedagogy and psychology. To sum up, the outcomes confirm that the lexical and semantic characteristics of specialized language underline social aspects of communication between experts (and public too). Further, we intend to elaborate words belonging to other spheres of professional communication (medicine, sociology, etc.) in the framework of the corpus examination.

Integration and analysis of geological, geochemical and remote sensing data of south of Neyshabur using principal component analysis

Lack of the existence of known mineral prospects in the preliminary stages of mineral exploration is the main problem of data-driven mineral potential modeling methods. On the other hand, applying the expert’s knowledge and judgment in different stages of mineral potential modeling, is the main difficulty of knowledge-driven mineral potential modeling methods. In addition, other difficulties in these methods can be mentioned such as determination of important variables, weighting to various classes of maps or information layers, and so on. Hence, the accuracy of the results of the knowledge-driven modeling methods is highly dependent on the amount of knowledge and experience of the expert. In this study, the principal component analysis (PCA) has been introduced as a knowledge-driven method with the least reliance on the expert’s knowledge for mineral potential modeling. In this method, the expert’s knowledge is only used in the interpretation of the results obtained from the modeling, and is not considered in the first stages of mineral potential modeling and definition of the conceptual model. In the introduced method, the interpretation of the results is conducted based on the positive and negative coefficients of variables in the eigenvalues table. Using these coefficients, it is determined that each principal component (PC) is associated with what type of mineralization. An advantage of this introduced method is to identify various types of mineralization in the area of interest using just one modeling effort. In order to evaluate the efficiency of this method, a region including two geological maps of Kadkan and Shamkan in the south of Neyshabur, northeast of Iran was selected. Two mineralization types including podiform chromite and epithermal gold-antimony mineralization types have been identified using the proposed method that presents more precise results than those of conventional univariate and multivariate geochemical studies.

An improved pattern-based prediction model for a class of industrial processes

For the problem of simplifying pattern-based modeling procedures, an improved pattern-based modeling method is put forward via pattern classification for a class of complex processes. It is a pure data-driven modeling method using statistical attributes of the processes. At the beginning of the paper, a method of system dynamics description based on pattern moving is introduced. Then, an improved method of pattern-moving-based prediction modeling is put forward, and it simplifies the pattern-moving-based modeling method by integrating an initial model and a classification mapping. It consists of two parts: system pattern construction and pattern classification. And a constructive classification neural network (CCNN) is designed to describe system dynamics by classifying the system pattern, and its generalization is discussed. Finally, simulations using data of an actual production process demonstrate the feasibility of the proposed modeling method, and the effectiveness of the CCNN is verified using comparison experiments.

Strategies for 3D Modelling of Buildings from Airborne Laser Scanner and Photogrammetric Data Based on Free-Form and Model-Driven Methods: The Case Study of the Old Town Centre of Bordeaux (France)

This paper presents a data-driven free-form modelling method dedicated to the parametric modelling of buildings with complex shapes located in particularly valuable Old Town Centres, using Airborne LiDAR Scanning (ALS) data and aerial imagery. The method aims to reconstruct and preserve the input point cloud based on the relative density of the data. The method is based on geometric operations, iterative transformations between point clouds, meshes, and shape identification. The method was applied on a few buildings located in the Old Town Centre of Bordeaux (France). The 3D model produced shows a mean distance to the point cloud of 0.058 m and a standard deviation of 0.664 m. In addition, the incidence of building footprint segmentation techniques in automatic and interactive model-driven modelling was investigated and, in order to identify the best approach, six different segmentation methods were tested. The segmentation was performed based on the footprints derived from Digital Surface Model (DSM), point cloud, nadir images, and OpenStreetMap (OSM). The comparison between the models shows that the segmentation that produces the most accurate and precise model is the interactive segmentation based on nadir images. This research also shows that in modelling complex structures, the model-driven method can achieve high levels of accuracy by including an interactive editing phase in building 3D models.

Exploring the Direct and Indirect Use of ICT Measurements in DODME (Dynamic OD Matrix Estimation)

The estimation of the network traffic state, its likely short-term evolution, the prediction of the expected travel times in a network, and the role that mobility patterns play in transport modeling is usually based on dynamic traffic models, whose main input is a dynamic origin–destination (OD) matrix that describes the time dependencies of travel patterns; this is one of the reasons that have fostered large amounts of research on the topic of estimating OD matrices from the available traffic information. The complexity of the problem, its underdetermination, and the many alternatives that it offers are other reasons that make it an appealing research topic. The availability of new traffic data measurements that were prompted by the pervasive penetration of information and communications technology (ICT) applications offers new research opportunities. This study focused on GPS tracking data and explored two alternative modeling approaches regarding how to account for this new information to solve the dynamic origin–destination matrix estimation (DODME) problem, either including it as an additional term in the formulation model or using it in a data-driven modeling method to propose new model formulations. Complementarily, independently of the approach used, a key aspect is the quality of the estimated OD, which, as recent research has made evident, is not well measured by the conventional indicators. This study also explored this problem for the proposed approaches by conducting synthetic computational experiments to control and understand the process.

Dynamic Equivalence Modeling for Microgrid Cluster by Using Physical-Data-Driven Method

In practical application of microgrid cluster, the lack of full detailed information cause the failure of dynamic modeling. Although some data-driven black-box modeling method can tackle this problem, insufficient usage of prior known physical information may reduce the modeling accuracy. To tackle this challenge, a hybrid physical-data-driven method is proposed for the dynamic behavior modeling of microgrid cluster. Motivated by the equivalence of recurrent neural network (RNN) and differential equations, the differential-algebraic equations (DAEs) of unknown part are represented by gate recurrent unit (GRU) based neural network. The DAEs of prior known physical stage are embedded into the proposed neural network, which avoid unnecessary model training of prior known section and improving the modeling efficiency. At first, the basic idea of RNN based dynamic modeling is explained. Then, the modeling guidelines including data preparation and parameter design are suggested. Finally, the effectiveness of the proposed method is confirmed by a test system formed by three microgrids under grid fault and operating point changing conditions.

A transfer-learning based energy consumption modeling method for industrial robots

Industrial robots have been widely utilized in factories to replace or help employees performing tasks such as handling, welding, and assembly. With the increasing breadth of installation and deployment of industrial robots, it is critical to predict and optimize their energy consumption in order to assure their environmentally friendly characteristics. While the data-driven modeling method based on multi-layer perception has been demonstrated to be a viable way for revealing quantitative relationships between operating parameters and energy consumption for industrial robots regardless of physical process or case-specific information, there are still some limitations in terms of prediction range, modeling efficiency, and sample size requirement. To accelerate model rebuilding and improve model accuracy without collecting a large number of samples simultaneously, this paper proposes a transfer-learning-based model creation method for energy consumption of industrial robots, where the associated operating parameters are analyzed, and structure adjustment strategies for multi-layer perception are planned for various industrial robot systems, then, schemes for the reuse of parameters in well-trained networks on source domain and the fine-tunning of multi-layer perception model are formulated to efficiently build the accurate energy consumption model on target domains. Experiments are conducted on two distinct robot systems, the Epson C4 and the Siasun SR10C, with nonidentical operating parameter structures and sample sizes. The results demonstrate that transfer learning models with fine-tuning strategy outperform previous data-driven modeling methods in terms of model accuracy and modeling efficiency in both of these situations.

Synergisms of machine learning and constraint-based modeling of metabolism for analysis and optimization of fermentation parameters.

Recent noteworthy advances in developing high-performing microbial and mammalian strains have enabled the sustainable production of bio-economically valuable substances such as bio-compounds, biofuels, and biopharmaceuticals. However, to obtain an industrially viable mass-production scheme, much time and effort are required. The robust and rational design of fermentation processes requires analysis and optimization of different extracellular conditions and medium components, which have a massive effect on growth and productivity. In this regard, knowledge- and data-driven modeling methods have received much attention. Constraint-based modeling (CBM) is a knowledge-driven mathematical approach that has been widely used in fermentation analysis and optimization due to its ability to predict the cellular phenotype from genotype through high-throughput means. On the other hand, machine learning (ML) is a data-driven statistical method that identifies the data patterns within sophisticated biological systems and processes, where there is inadequate knowledge to represent underlying mechanisms. Furthermore, ML models are becoming a viable complement to constraint-based models in a reciprocal manner when one is used as a pre-step of another. As a result, a more predictable model is produced. This review highlights the applications of CBM and ML independently and the combination of these two approaches for analyzing and optimizing fermentation parameters.

Integrated spatiotemporal modeling and mixed-integer approximate dynamic programming for ASP flooding

Alkali–surfactant–polymer (ASP) flooding holds a critical position in oil development. This paper is concerned with the distributed parameter system (DPS) modeling and development scheme optimization of ASP flooding. Owing to the neglect of coupling among three-dimensional (3-D) spatial information, most of the existing data-driven modeling methods fail to provide a high-precision model for this type of strongly nonlinear and spatiotemporal coupling 3-D DPS. To overcome this shortcoming, an integrated spatiotemporal modeling method (ISM) is proposed, which reconstructs temporal–spatial relationship to avoid spatial mapping distortion based on canonical polyadic decomposition (CPD), and a spatiotemporal MSSO-LS-SVM algorithm is presented to compensate the nonlinear dynamics and decrease truncation error. Moreover, the synchronous optimization of ASP injection strategy and oil wells switching scheduling is an urgent industry requirement of intelligent oilfield decision system, which can be described a mixed-integer optimal control/dynamic programming problem. This paper first puts forward a mixed-integer approximate dynamic programming (MIADP) algorithm, dedicated for solving this class of optimal control problem with integer control subset requirements. In the MIADP, a parallel mixed-integer policy improvement framework is developed, where (i) the continuous control weight updates by the temporal difference (TD) error, (ii) the quantum evolution mechanism is designed to learn the integer control, and (iii) the two parts share information interactively and iterate to approach the optimal mixed-integer policy. At last, a case study on ASP flooding is taken, and the superior performance metrics illustrate the effectiveness of the proposed ISM method and MIADP algorithm.

Data-driven fault tolerant predictive control for temperature regulation in data center with rack-based cooling architecture

Recently increasing amount of data centers (DCs) have been constructed for the development of information and communications technology (ICT). Meanwhile, more energy consumption is required to support the operation of DCs, where nearly thirty percent of the consumption is used for cooling system. Therefore, this paper studies an efficient cooling architecture named rack mountable cooling unit (RMCU) for DCs, and provides a novel data-driven fault tolerant predictive controller to regulate the temperature accordingly. In order to circumvent the complicated physics modeling in DC, a data-driven modeling method is developed. In this method, multiple local linear models in form of auto-regressive exogenous (ARX) model are selected to describe the strongly nonlinear system. Besides, the multiple sub-models are identified through partial least square (PLS) and corresponding training data is partitioned by fuzzy c-means (FCM). Then, on the basis of developed data-driven model, a predictive controller considering actuator faults is designed to manage the temperature in DC. Finally, real experimental results are presented and demonstrate the superior performance of our proposed controller.

Bounded Cost Path Planning for Underwater Vehicles Assisted by a Time-Invariant Partitioned Flow Field Model.

A bounded cost path planning method is developed for underwater vehicles assisted by a data-driven flow modeling method. The modeled flow field is partitioned as a set of cells of piece-wise constant flow speed. A flow partition algorithm and a parameter estimation algorithm are proposed to learn the flow field structure and parameters with justified convergence. A bounded cost path planning algorithm is developed taking advantage of the partitioned flow model. An extended potential search method is proposed to determine the sequence of partitions that the optimal path crosses. The optimal path within each partition is then determined by solving a constrained optimization problem. Theoretical justification is provided for the proposed extended potential search method generating the optimal solution. The path planned has the highest probability to satisfy the bounded cost constraint. The performance of the algorithms is demonstrated with experimental and simulation results, which show that the proposed method is more computationally efficient than some of the existing methods.

Data-driven cloud simulation architecture for automated flexible production lines: application in real smart factories

In recent years, more manufacturing enterprises are building automated flexible production lines (AFPLs) to satisfy the dynamic and diversified demand. Currently, static planning methods can hardly meet the requirements of the dynamic resource allocation for AFPLs. The technologies of the digital twin can help solve dynamic problems. Therefore, we propose a data-driven cloud simulation architecture for AFPLs in smart factories. First, we design a cloud simulation platform as the architecture foundation. Second, we use the data-driven modelling and simulation method to achieve automated modelling. Third, we implement the system on the cloud using Java, MySQL, and the Anylogic platform, and verify the efficiency of the proposed method by experiments in the real workshop of a 3C (Computer, Communication, Consumer electronics) company. The experimental results show the proposed architecture can support the real-time resource allocation decisions to maximise the throughput in AFPLs. This paper makes contributions by proposing an architecture realising automatic modelling and data-driven simulation first in the cloud simulation environment, and filling the gap of dynamic resource allocation in the research of AFPLs.

Data‐driven models for short‐term ocean wave power forecasting

In order to integrate wave farms into the grid, the power from wave energy converters (WEC) must be forecasted. This study presents a novel data-driven modelling (DDM) method to predict very short-term (15 min–4 h) and short-term (0–72 h) power generation from a WEC. The model is tested using data from an oscillating body converter. Several other methods are tested as well. These include support vector machines (SVM), neural networks (NN), and recurrent neural networks (RNN). Of these, the best is the long-short-term memory (LSTM) network, which is trained and updated on observed values. The experiments demonstrate both the SVM and NN forecast well. However, the proposed deep learning models predict them more accurately. The models work well over short horizons. At horizons longer than three days, accuracy deteriorates, and the models cannot fit the data well.

Artificial Intelligence in Optical Communications: From Machine Learning to Deep Learning

Techniques from artificial intelligence have been widely applied in optical communication and networks, evolving from early machine learning (ML) to the recent deep learning (DL). This paper focuses on state-of-the-art DL algorithms and aims to highlight the contributions of DL to optical communications. Considering the characteristics of different DL algorithms and data types, we review multiple DL-enabled solutions to optical communication. First, a convolutional neural network (CNN) is used for image recognition and a recurrent neural network (RNN) is applied for sequential data analysis. A variety of functions can be achieved by the corresponding DL algorithms through processing the different image data and sequential data collected from optical communication. A data-driven channel modeling method is also proposed to replace the conventional block-based modeling method and improve the end-to-end learning performance. Additionally, a generative adversarial network (GAN) is introduced for data augmentation to expand the training dataset from rare experimental data. Finally, deep reinforcement learning (DRL) is applied to perform self-configuration and adaptive allocation for optical networks.

Implicit generalized predictive control method for high-speed train

The running process of the train which can high speed running is a typical nonlinear process, and with the higher and higher velocity of the train which can be driven extremely fast, the characteristics of the nonlinear will be stronger than before, which leads to higher requirements for the design of the “Automatic Train Operation” driving system. Aiming at the characteristics of high-speed train running speed fast, complex operation process and strong nonlinearity, the mathematical model establishment and the speed tracking control method are studied. This article uses a data-driven modeling method. Firstly, the actual operation data of high-speed trains based on traction and braking forces are divided into two parts: traction and braking. Then, the linear models of the traction and braking models of the train, which can be driven very fast, are established respectively. These two models adopt the recursive least square method. Then use implicit generalized predictive control to achieve tracking control of high-speed train speed. The simulation results based on the operating process data of CRH380AL is given that this method can make the requirements that belong to high-speed train operation.

A Novel Data-Driven Modeling and Control Design Method for Autonomous Vehicles

This paper presents a novel modeling method for the control design of autonomous vehicle systems. The goal of the method is to provide a control-oriented model in a predefined Linear Parameter Varying (LPV) structure. The scheduling variables of the LPV model through machine-learning-based methods using a big dataset are selected. Moreover, the LPV model parameters through an optimization algorithm are computed, with which accurate fitting on the dataset is achieved. The proposed method is illustrated on the nonlinear modeling of the lateral vehicle dynamics. The resulting LPV-based vehicle model is used for the control design of path following functionality of autonomous vehicles. The effectiveness of the modeling and control design methods through comprehensive simulation examples based on a high-fidelity simulation software are illustrated.