Exploring the value of using data: a case study of continuous improvement through data warehousing

The mining industry is adapting data-rich technology. Data management processes and infrastructure require additional investment which, can be difficult to justify. A valuation model for data management infrastructure would necessitate modelling the ‘use’ of data and quantifying the value and the likelihood of success of improvement initiatives. A model is presented which can assess the impacts of a data-driven initiative. The initiatives are shown to be the foundation of improved shift coordination resulting in significant and sustained operational improvements. This paper aims to illustrate the value of using data from source systems in mining operations.     

Multivariate GP-VAR models for robust structural identification under operational variability

While the concept of structural monitoring has been around for a number of decades, it remains under-exploited in practice. A main driver for this shortcoming lies in the difficulty to robustly and autonomously interpret the information that is extracted from dynamic data. This hindrance in properly deciphering the collected information may be attributed to the uncertainty that is inherent in i) the finite set of measured data, ii) the models employed for capturing the manifested dynamics, and more importantly, iii) the susceptibility of these systems to variations in Environmental and Operational Parameters (EOPs). In previous work of the authors, a Gaussian Process (GP) time-series approach has been introduced, which serves as a hierarchical input–output method to account for the influence of EOPs on structural response. This in turn enables a robust structural identification. In this scheme, the short-term dynamics are modeled by means of linear-in-the-parameters time-series models, while EOV dependence – acting on a long-term time scale – is achieved via GP regression of the model coefficients on measured EOPs. This work corresponds to a further advancement on this modeling approach, corresponding to its generalization to the vector response case. Particularly, the problem of global identification here is solved via an Expectation–Maximization algorithm tailored to the GP time-series model structure. Moreover, an EOP-dependent innovations covariance matrix is integrated in the model, which helps to capture variation in the vibration power. The resulting model does not only have the capability to represent the long-term response of a structure under variable EOPs, but also facilitates the enhanced tracking of modal quantities in contrast to traditional operational modal analysis techniques. The proposed approach is exemplified on the identification of the vibration response of a simulated wind turbine blade at different points along the blade axis in the flap-wise direction, under variability of both the acting wind speeds and ambient temperatures.     

Deep Learning in Sheet Metal Bending With a Novel Theory-Guided Deep Neural Network

Sheet metal forming technologies have been intensively studied for decades to meet the increasing demand for lightweight metal components. To surmount the springback occurring in sheet metal forming processes, numerous studies have been performed to develop compensation methods. However, for most existing methods, the development cycle is still considerably time-consumptive and demands high computational or capital cost. In this paper, a novel theory-guided regularization method for training of deep neural networks (DNNs), implanted in a learning system, is introduced to learn the intrinsic relationship between the workpiece shape after springback and the required process parameter, e.g., loading stroke, in sheet metal bending processes. By directly bridging the workpiece shape to the process parameter, issues concerning springback in the process design would be circumvented. The novel regularization method utilizes the well-recognized theories in material mechanics, Swift’s law, by penalizing divergence from this law throughout the network training process. The regularization is implemented by a multi-task learning network architecture, with the learning of extra tasks regularized during training. The stress-strain curve describing the material properties and the prior knowledge used to guide learning are stored in the database and the knowledge base, respectively. One can obtain the predicted loading stroke for a new workpiece shape by importing the target geometry through the user interface. In this research, the neural models were found to outperform a traditional machine learning model, support vector regression model, in experiments with different amount of training data. Through a series of studies with varying conditions of training data structure and amount, workpiece material and applied bending processes, the theory-guided DNN has been shown to achieve superior generalization and learning consistency than the data-driven DNNs, especially when only scarce and scattered experiment data are available for training which is often the case in practice. The theory-guided DNN could also be applicable to other sheet metal forming processes. It provides an alternative method for compensating springback with significantly shorter development cycle and less capital cost and computational requirement than traditional compensation methods in sheet metal forming industry.      

Joint input shaping and feedforward for point-to-point motion: Automated tuning for an industrial nanopositioning system

Feedforward control can effectively compensate for the servo error induced by the reference signal if it is tuned appropriately. This paper aims to introduce a new joint input shaping and feedforward parametrization in iterative feedforward control. Such a parametrization has the potential to significantly improve the performance for systems executing a point-to-point reference trajectory. The proposed approach enables an efficient optimization procedure with global convergence. A simulation example and an experimental validation on an industrial motion system confirm (i) the performance improvement obtained by means of the joint input shaping and feedforward parametrization compared to pre-existing results, and (ii) the efficiency of the proposed optimization procedure.     

Multiple optimized online support vector regression for adaptive time series prediction

Data-driven prognostics based on sensor or historical test data have become appropriate prediction means in prognostics and health management (PHM) application. However, most traditional data-driven prognostics methods are off-line which would be seriously limited in many PHM systems needed on-line predicting or real-time processing. Furthermore, even in some on-line prediction algorithms such as Online Support Vector Regression (Online SVR) and Incremental learning algorithm, there are conflicts and trade-offs between prediction efficiency and accuracy. Therefore, in different PHM applications, prognostics algorithms should be on-line, flexible and adaptive to balance the prediction efficiency and accuracy. An on-line adaptive data-driven prognostics strategy is proposed with five various optimized on-line prediction algorithms based on Online SVR. These five algorithms are improved with kernel combination and sample reduction to realize higher precision and efficiency. These algorithms can achieve more accurate results by data pre-processing and model optimization, moreover, faster operating speed and lower computational complexity can be obtained by optimization of training process with on-line data reduction. With these different improved Online SVR methods, varies of prediction with different precision and efficiency demands could be fulfilled by an adaptive strategy. To evaluate the proposed prognostics strategy, we have executed simulation experiments with Tennessee Eastman (TE) process. In addition, the prediction strategies are also applied and evaluated by traffic mobile communication data from China Mobile Communications Corporation Heilongjiang Co., Ltd. Experiments and test results prove its effectiveness and confirm that the algorithms can be effectively applied to the on-line status prediction with increased performance in both precision and efficiency.     

Data-driven digital twin technology for optimized control in process systems

Due to the installation of various apparatus in process industries, both factors of complex structures and severe operating conditions could result in higher accident frequencies and maintenance challenges. Given the importance of security in process systems, this paper presents a data-driven digital twin system for automatic process applications by integrating virtual modeling, process monitoring, diagnosis, and optimized control into a cooperative architecture. For unknown model parameters, the adaptive system identification is proposed to model closed-loop virtual systems and residual signals with fault-free case data. Performance indices are improved to make the design of robust monitoring and diagnosis system to identify the apparatus status. Soft-sensor, parameterization control, and model-matching reconfiguration are ameliorated and incorporated into the optimized control configuration to guarantee stable and safe control performance under apparatus faults. The effectiveness and performance of the proposed digital twin system are evaluated by using different simulations on the Tennessee Eastman benchmark process in the presence of realistic fault scenarios.     

Evolutionary Design of Nickel-Based Superalloys Using Data-Driven Genetic Algorithms and Related Strategies

Data-driven models were constructed for the mechanical properties of multi-component Ni-based superalloys, based on systematically planned, limited experimental data using a number of evolutionary approaches. Novel alloy design was carried out by optimizing two conflicting requirements of maximizing tensile stress and time-to-rupture using a genetic algorithm-based multi-objective optimization method. The procedure resulted in a number of optimized alloys having superior properties. The results were corroborated by a rigorous thermodynamic analysis and the alloys found were further classified in terms of their expected levels of hardenabilty, creep, and corrosion resistances along with the two original objectives that were optimized. A number of hitherto unknown alloys with potential superior properties in terms of all the attributes ultimately emerged through these analyses. This work is focused on providing the experimentalists with linear correlations among the design variables and between the design variables and the desired properties, non-linear correlations (qualitative) between the design variables and the desired properties, and a quantitative measure of the effect of design variables on the desired properties. Pareto-optimized predictions obtained from various data-driven approaches were screened for thermodynamic equilibrium. The results were further classified for additional properties.     

Iterative motion feedforward tuning: A data-driven approach based on instrumental variable identification

Feedforward control can significantly enhance the performance of motion systems through compensation of known disturbances. This paper aims to develop a new procedure to tune a feedforward controller based on measured data obtained in finite time tasks. Hereto, a suitable feedforward parametrization is introduced that provides good extrapolation properties for a class of reference signals. Next, connections with closed-loop system identification are established. In particular, instrumental variables, which have been proven very useful in closed-loop system identification, are selected to tune the feedforward controller. These instrumental variables closely resemble traditional engineering tuning practice. In contrast to pre-existing approaches, the feedforward controller can be updated after each task, irrespective of noise acting on the system. Experimental results confirm the practical relevance of the proposed method.     

Transient identification in nuclear power plants: A review

A transient is defined as an event when a plant proceeds from a normal state to an abnormal state. In nuclear power plants (NPPs), recognizing the types of transients during early stages, for taking appropriate actions, is critical. Furthermore, classification of a novel transient as “don't know”, if it is not included within NPPs collected knowledge, is necessary. To fulfill these requirements, transient identification techniques as a method to recognize and to classify abnormal conditions are extensively used. The studies revealed that model-based methods are not suitable candidates for transient identification in NPPs. Hitherto, data-driven methods, especially artificial neural networks (ANN), and other soft computing techniques such as fuzzy logic, genetic algorithm (GA), particle swarm optimization (PSO), quantum evolutionary algorithm (QEA), expert systems are mostly investigated. Furthermore, other methods such as hidden Markov model (HMM), and support vector machines (SVM) are considered for transient identification in NPPs. By these modern techniques, NPPs safety, due to accidents recognition by symptoms rather than events, is improved. Transient identification is expected to become increasingly important as the next generation reactors being designed to operate for extended fuel cycles with less operators' oversight. In this paper, recent studies related to the advanced techniques for transient identification in NPPs are presented and their differences are illustrated.