Dynamic Batch Process Research Articles

BatOpt: Optimizing GPU-Based Deep Learning Inference Using Dynamic Batch Processing

BatOpt: Optimizing GPU-Based Deep Learning Inference Using Dynamic Batch Processing

Data-Driven Soft Sensing for Batch Processes Using Neural Network-Based Deep Quality-Relevant Representation Learning

Soft sensors provide a means to reliably estimate unmeasurable variables, thereby playing a prevalent role in formulating closed-loop control in batch processes. In soft sensor development, enhancing quality-relevant information and eliminating quality-irrelevant information are important. This study proposes a neural network-based deep quality-relevant representation learning approach to improve the soft sensing performance in dynamic batch processes. The structure of a deep neural network is optimized in a layer-by-layer manner. First, given the generally abundant predictor variables, the maximal relevanceminimal redundancy criterion is used to optimize the input layer, select the most beneficial variables, and eliminate modeling redundancy. Second, mutual information-based quality-relevant representation selection is performed in the middle layers to enhance the quality-relevant information and eliminate the influence of irrelevant representations. Third, deep quality-relevant representations are extracted, and a soft sensor model is developed. The proposed method is tested on a fed-batch penicillin fermentation process and an industrial injection molding process. Lastly, it shows that the proposed method outperforms several state-of-the-art approaches, thereby confirming its effectiveness.

Dynamic Batch Process Monitoring Based on Time-Slice Latent Variable Correlation Analysis.

Batch processes are generally characterized by complex dynamics and remarkable data collinearity, thereby rendering the monitoring of such processes necessary but challenging. This paper proposes a data-driven time-slice latent variable correlation analysis-based model predictive fault detection framework to ensure accurate fault detection in dynamic batch processes. The three-way batch process data are first unfolded into the two-way time slice. For each single time slice, process data are mapped to both major latent variables and residual subspaces to deal with the variable-wise data collinearity and extract dominant data information. A measurement status is then determined with a canonical correlation analysis of the major latent variables and correlated variables, using both the time and batch perspectives. Prediction-based residuals are generated, which provide the basis for identifying the property of faults detected, namely, static or dynamic. Based on experiments using a simulated penicillin production and an industrial inject molding process, the proposed monitoring scheme has been proven feasible and effective.

Batch process monitoring based on global enhanced multiple neighborhoods preserving embedding

Batch process generally has varying dynamic characteristic that causes low fault detection rate and high false alarm rate, and it is necessary and urgent to monitor batch process. This paper proposes a global enhanced multiple neighborhoods preserving embedding based fault detection strategy for dynamic batch process. Firstly, the angle neighbor is defined and selected to compensate for the insufficient expression for the spatial similarity of samples only by using the distance neighbor, and the time neighbor is introduced to describe the time correlations between samples. These three types of neighbors can fully characterize the similarity of the samples in time and space. Secondly, considering the minimum reconstruction error and the order information of three types of neighbors, an enhanced objective function is constructed to prevent the loss of order information when neighborhood preserving embedding (NPE) calculates the reconstruction weights. Furthermore, the enhanced objective function and a global objective function are organically combined to extract both global and local features, to describe process dynamics and visualize process data in a low-dimensional space. Finally, a monitoring index based on support vector data description is constructed to eliminate adverse effects of non-Gaussian data for monitoring performance. The advantages of the proposed method over principal component analysis, neighborhood preserving embedding, dynamic principal component analysis and time NPE are demonstrated by a numerical example and the penicillin fermentation process simulation.

Proactive Fragmentation Management Scheme Based on Crosstalk-Avoided Batch Processing for Spectrally-Spatially Elastic Optical Networks

Fragmentation with crosstalks is the major obstacle in spectrally-spatially elastic optical networks, which suppresses resource utilization while degrading the quality-of-transmission. To overcome this issue, this article proposes, for the first time, a proactive fragmentation management scheme based on batch processing while satisfying both inter-core and inter-mode crosstalks to enhance resource utilization. The proposed scheme adopts a batch processing method to create batches of lightpath requests received within a time threshold to utilize spectrum resources effectively. In batch processing, lightpath requests are prioritized based on the number of links in their routes and required slots. To maintain fairness in batch processing, when any request is rejected, the proposed scheme triggers a procedure that gives an equal opportunity to all arriving requests within the threshold, irrespective of numbers of hops and requested capacities, for allocation. We formulate the static batch processing of lightpath requests (SBPLR) as an integer linear programming (ILP) problem. We prove that SBPLR is an NP-Complete problem. We introduce a heuristic solution when ILP is intractable. To serve lightpath requests in each batch while avoiding inter-core and inter-mode crosstalks, we develop a core-mode-spectrum allocation algorithm. We present a dynamic batch processing based fragmentation management approach. Numerical results indicate that the proposed scheme outperforms the benchmark schemes.

Multiphase two-dimensional time-slice dynamic system for batch process monitoring

In this work, a multiphase 2D time-slice dynamic system approach has been proposed to characterize the batch-wise and variable-wise dynamics of batch processes simultaneously. First, in each phase, the 2D latent representation has been developed for time slice matrices so as to extract informative features from both batch-wise and variable-wise variations. Afterward, the first-order Markov chains are introduced for capturing the underlying 2D latent dynamics across time-slice evolutions. In this way, batch process dynamic behaviors in each working phase can be captured from both single batch run and batch-to-batch evolutions, while process noise can also be explained in the probabilistic modeling framework. Last, in order to carry out online monitoring with the proposed 2D dynamic system, an effective 2D online-monitoring mechanism is established. The feasibility of the dynamic batch process monitoring method is validated on the fed-batch penicillin fermentation benchmark.

Just‐in‐time learning–multiple subspace support vector data description used for non‐Gaussian dynamic batch process monitoring

AbstractBatch process data are time‐varying dynamic and non‐Gaussian distributed. In addition, for multivariate statistical process monitoring, their variability can be overwhelmed when considering local variability behavior. To address the abovementioned issues, an improved batch process monitoring approach is presented that integrates just‐in‐time learning and multiple subspace support vector data description (JITL‐MSSVDD). A new multiple subspace segmentation method is proposed that classifies a contribution array that is calculated on the mixing matrix of independent component analysis (ICA). Offline, the variable subspace segmentation rule can be obtained from the proposed method. The subspace monitoring models can reduce the risk of the variability being overwhelmed. Online, local modeling samples are collected through JITL, which can reduce the impact of the time‐varying dynamic on modeling accuracy. Then, in accordance with the variable subspace segmentation rule, which is obtained offline, MSSVDD models are constructed to solve non‐Gaussian problems. The advantage of the proposed JITL‐MSSVDD is demonstrated through the standard test model fed‐batch penicillin fermentation process.

Challenges of Processes Simulation with Dynamic Batch Processing Activities

Business Process Management (BPM) often encompasses process improvements based on Information and Communication Technologies (ICT). As a tool for estimating optimization parameters, business processes simulation can be used. Commercially available BPM tools with BPMN 2.0 accordance, have built-in simulation features which vary in the extent of supporting advanced process simulation options. In this paper, we address dynamic batch processing activities as a challenge for BPM tools with simulation features and their dynamic impact in executing activities. Further on, basic process scenarios examples were used for testing and evaluating the capabilities of selected BPM tools especially in relation to Time Driven Activity Based Costing model (TD ABC model). Closing part of the paper deals with significant limitations of BPM tools in supporting batch processing activities and the comparison to TD ABC model simulation as temporary solution for these problems.

Batch process monitoring based on self-adaptive subspace support vector data description

Inherent time-varying dynamics, which is a general characteristic of batch processing, causes two problems in data-driven batch process monitoring methods: (1) changes in data trajectory and (2) changes in correlation between variables along time. These problems can be solved by employing monitoring methods based on moving time window technology. However, correlation behaviors between variables in dynamic batch processing are complex. As a consequence, traditional monitoring methods may fail to detect faults. Complex correlation behaviors of batch processing can be learned by placing variables with similar variation information in the same subspace and faults may be detected. In this study, a self-adaptive subspace support vector data description (SASSVDD) is proposed. Two-time unfolding three-dimensional data technology and moving time window technology are used to obtain modeling data. An online subspace is then constructed by using sensitive variables, which may highly yield variation information, and non-sensitive variables, which likely contain variation information and exhibit a higher correlation with sensitive variables. Support vector data description is applied as the subspace monitoring method. The availability of SASSVDD is verified through the fed-batch penicillin fermentation.

Weak fault monitoring method for batch process based on multi-model SDKPCA

In industrial manufacturing, most batch processes have the dynamic and nonlinear features in nature. To ensure both quality consistency of the manufactured products and safe operation of this kind of batch process, a number of multivariate statistical analyses, including multiway principal component analysis (MPCA), batch dynamic kernel principal component analysis (BDKPCA), have been developed in recent years. However, these methods can't effectively detect the weak faults due to large fluctuations in the initial conditions, because the weak faults are submerged to the fluctuations in the poor initial conditions. In order to improve the performance of the weak fault detection, a new nonlinear dynamic batch process monitoring method, called multi-model single dynamic kernel principal component analysis (M-SDKPCA), is proposed in this paper. The multi-model methodology is based on BDKPCA. The method firstly integrates kernel PCA (KPCA) and auto-regressive moving average exogenous (ARMAX) time series model for each batch data at each stage to build SDKPCA. Then hierarchical clusters are obtained through load matrix similarity among SDKPCA models. At different stages, multiple model structures are constructed along with the variation of the cluster number. The monitoring method proposed in this paper was applied to fault detection for benchmark of fed-batch penicillin production. In both off-line analysis and on-line batch monitoring, the proposed approach shows better performance than MKPCA and BDKPCA.

Multiway discrete hidden Markov model‐based approach for dynamic batch process monitoring and fault classification

AbstractA new multiway discrete hidden Markov model (MDHMM)‐based approach is proposed in this article for fault detection and classification in complex batch or semibatch process with inherent dynamics and system uncertainty. The probabilistic inference along the state transitions in MDHMM can effectively extract the dynamic and stochastic patterns in the process operation. Furthermore, the used multiway analysis is able to transform the three‐dimensional (3‐D) data matrices into 2‐D measurement‐state data sets for hidden Markov model estimation and state path optimization. The proposed MDHMM approach is applied to fed‐batch penicillin fermentation process and compared to the conventional multiway principal component analysis (MPCA) and multiway dynamic principal component analysis (MDPCA) methods in three faulty scenarios. The monitoring results demonstrate that the MDHMM approach is superior to both the MPCA and MDPCA methods in terms of fault detection and false alarm rates. In addition, the supervised MDHMM approach is able to classify different types of process faults with high fidelity. © 2011 American Institute of Chemical Engineers AIChE J, 2012

Improved two-dimensional dynamic batch process monitoring with support vector data description

AbstractFor dynamic batch process monitoring, a two-dimensional dynamic modeling framework has recently been formulated, which is based on a two-dimensional autoregressive model and the principal component analysis (PCA) method. Different from traditional dynamic batch process monitoring, the two-dimensional method can monitor both within batch and batch-to-batch dynamic information of the process data. However, this PCA-related method has two main restrictions, which may render poor monitoring performance in practice. First, it is under the assumption that the distribution of the process data is Gaussian. Second, the correlations between different process variables are assumed to be linear with each other. Unfortunately, both of these two assumptions are difficult to satisfy in batch processes. In this paper, support vector data description (SVDD) is incorporated into the two-dimensional modeling framework, which has no Gaussian limitation of the data, and can also model the nonlinear relationship between process variables. For dynamic batch process monitoring, a distance based statistic is proposed. Based on results of a simulation case study, the monitoring performance has been improved.

Statistical Monitoring and Fault Diagnosis of Batch Processes Using Two-Dimensional Dynamic Information

Two-dimensional (2D) dynamics widely exist in batch processes, which inspirit research efforts to develop corresponding monitoring schemes. Recently, two-dimensional dynamic principal component analysis (2D-DPCA) has been proposed to model and monitor such 2D dynamic batch processes, in which support region (ROS) determination is a key step. A proper ROS ensures modeling accuracy, monitoring efficiency, and reasonable fault diagnosis. The previous ROS determination method is practicable in many situations but still has certain limitations, as discussed in this paper. To overcome these shortcomings, a 2D-DPCA method with an improved ROS determination procedure is developed, by considering variable partial correlations and performing iterative stepwise regressions. Such a procedure expands ROS batch by batch and is a generalization of the autoregressive (AR) model order selection to the 2D batch process cases. Simulations show that the proposed method extracts 2D dynamics more accurately and improves the monitoring and diagnosis performance of the 2D-DPCA model.

Multivariate statistical monitoring of two-dimensional dynamic batch processes utilizing non-Gaussian information

Dynamics are inherent characteristics of batch processes, and they may exist not only within a particular batch, but also from batch to batch. To model and monitor such two-dimensional (2D) batch dynamics, two-dimensional dynamic principal component analysis (2D-DPCA) has been developed. However, the original 2D-DPCA calculates the monitoring control limits based on the multivariate Gaussian distribution assumption which may be invalid because of the existence of 2D dynamics. Moreover, the multiphase features of many batch processes may lead to more significant non-Gaussianity. In this paper, Gaussian mixture model (GMM) is integrated with 2D-DPCA to address the non-Gaussian issue in 2D dynamic batch process monitoring. Joint probability density functions (pdf) are estimated to summarize the information contained in 2D-DPCA subspaces. Consequently, for online monitoring, control limits can be calculated based on the joint pdf. A two-phase fed-batch fermentation process for penicillin production is used to verify the effectiveness of the proposed method.

Multivariate statistical monitoring of multiphase two-dimensional dynamic batch processes

To ensure product quality and operation safety, multivariate statistical process control (MSPC) techniques have been applied to batch process monitoring. Batch processes have several important features which should be taken into consideration in statistical monitoring, such as two-dimensional (2D) dynamics along both time and batch axes and multiple operation phases within a batch. In this paper, a multiphase two-dimensional dynamic PCA (2D-DPCA) method is proposed to deal with these characteristics simultaneously. An iterative procedure is designed for phase division, region of support (ROS) determination for each phase and phase 2D-DPCA modeling. The transition regions from phase to phase are also identified and modeled. Then, the phase and transition 2D-DPCA models are utilized in online monitoring. A three-water-tank process with 2D dynamics and a simulated two-phase batch process are used as case studies, to show the effectiveness of the proposed method.

Subspace identification for two-dimensional dynamic batch process statistical monitoring

Dynamics are inherent characteristics of batch processes, which may be not only within a batch, but also from batch to batch. Two-dimensional dynamic principal component analysis (2-D-DPCA) method [Lu, N., Yao, Y., Gao, F., 2005. Two-dimensional dynamic PCA for batch process monitoring. A.I.Ch.E. Journal 51, 3300–3304] can model both kinds of batch dynamics, but may lead to the inclusion of large number of lagged variables and make the contribution plot difficult to read. To solve this problem, subspace identification technique is combined with 2-D-DPCA in this paper. The state space model of a 2-D batch process can be identified with canonical variate analysis (CVA) method based on the auto-determined support region (ROS). In 2-D-DPCA modeling, the utilization of state variables instead of lagged process variables reduces the number of variables and provides a clearer contribution plot for fault diagnosis.

Batch Process Monitoring in Score Space of Two-Dimensional Dynamic Principal Component Analysis (PCA)

Two-dimensional dynamic principal component analysis (2-D-DPCA) is a recent developed method for two-dimensional (2-D) dynamic batch process monitoring. However, it only utilizes residual information in fault detection and information in score space is wasted, which may compromise the monitoring efficiency. In this paper, 2-D multivariate score autoregressive (AR) filters are designed to remove the 2-D dynamics retained in score space and make the filtered scores obey certain statistical assumptions, so that the T2 statistic can be calculated reasonably for process monitoring. Simulation shows that using the filters enhances the monitoring efficiency while reducing the chances of false alarms and missed alarms.

Dynamic batch processing in workflows: Model and implementation

Dynamic batch processing commonly exists in production or business processes. Traditional workflow management systems (WfMSs) do not support dynamic batch processing. This paper applies dynamic batch processing to WfMSs. The proposed mechanism consists of a dynamic batch processing scheduling model as well as the design and implementation. Our experiment shows that the proposed dynamic batch processing mechanism is useful for and convenient to business processes.

Using Taguchi's Method and Orthogonal Function Approximation to Design Optimal Manipulated Trajectory in Batch Processes

A novel experimental scheme is proposed to integrate orthogonal function approximation with Taguchi's method (orthogonal array) for designing the optimal manipulated trajectory of a batch process. The orthogonal function approximation finds a set of orthonormal functions as the basis to represent the batch trajectory. The optimal trajectory can be obtained if the location of the coefficients of the orthonormal functions is properly adjusted in the function space. The Taguchi approach is used to design and analyze the effect of each coefficient on reaching the optimal objective (quality) function. Because the coefficients are implicitly related to the objective function, they simply vary over two levels in a systematic way, and they would be moved into the optimal design condition. A search procedure for the optimal design coefficients is also proposed. As opposed to model-based design, the proposed method utilizes the simplicity of Taguchi methods to determine the potentially available knowledge of dynamic batch processes. For its potential applications, this combinational method is demonstrated through two simulation case studies.

Dynamic Optimization in the Design and Scheduling of Multiproduct Batch Plants

Dynamic batch processing provides additional transient operating freedom, that can stretch the limits of profitability under strict market, facility, and time constraints. This paper incorporates dynamic processing conditions for products in a multiproduct batch plant, as opposed to fixing the process by recipes, in the broader context of equipment design, production planning, scheduling, and inventory considerations. The objective is a general function of fixed design costs, operating costs, production revenues, etc. Decisions include stage processing times for products, transient stage operating policies, continuous design parameters, production capacity, and production schedules. The infinite dimensional optimal control problem for each operation is solved using collocation over finite time elements (Cuthrell, J. E.; Biegler, L. T. Comp. Chem. Eng. 1989, 13, 49. Logsdon, J. S.; et al. Trans. Inst. Chem. Eng. 1990, 68A, 434). Scheduling, with its combinatorial complexity, is addressed in the scope of flowshop plants for specific transfer policies using the aggregated scheduling model in Birewar, D. B.; Grossmann, I. E. Ind. Eng. Chem. Res. 1989a, 28, 1333; Comp. Chem. Eng. 1989b, 13, 141.