On-line Process Monitoring Research Articles

Online monitoring of IGBT modules based on creating the non-interventional monitoring environment

Aiming at the health monitoring of IGBT modules based on the online measurement of junction temperature Tj and the saturated voltage drop VCE in the three-level neutral-point-clamped (3LNPC) inverter, this paper puts forward the idea of creating a suitable monitoring environment to inject monitoring current without affecting the normal operation of the inverter. Based on this method, not only the online monitoring can be performed at different moments, but also the monitoring duration can be selected according to monitoring requirements. Meanwhile, the online monitoring will not distort the load current of the inverter. In this paper, an online monitoring experiment platform has also been designed to realize the monitoring current injection, the saturated voltage drop VCE sampling, and the sampling data transmission. The results of simulation and experiment verify the safety and the non-intervention of the online monitoring process, and the data collected by DSP during monitoring is used to judge whether the bond wires of IGBT module are failed.

A Fabry-Perot temperature sensor sealed with thermo-sensitive polymer

A fiber Fabry-Perot (F-P) temperature sensor sealed with thermosensitive polymer has been proposed and demonstrated. The structure of fully enclosed quartz probe reduces the interference from ambient pressure and refractive index (RI). The automatic fabrication process of microflow injection and discharge welding is simple and improves the controllability and repeatability of the structural parameters. A demodulation method is proposed for expanding the measurement range of temperature. The results show that the sensor has good linear response with a sensitivity of −1.014 nm/°C in the range of 30 ∼ 70 °C. In addition, the interference spectrum doesn’t change with ambient RI. The sensor has a good application prospect in the field of biochemical process on-line monitoring, such as cell culture.

On-Line Real-Time Monitoring of a Rapid Enzymatic Oil Degumming Process: A Feasibility Study Using Free-Run Near-Infrared Spectroscopy.

Enzymatic degumming is a well established process in vegetable oil refinement, resulting in higher oil yield and a more stable downstream processing compared to traditional degumming methods using acid and water. During the reaction, phospholipids in the oil are hydrolyzed to free fatty acids and lyso-phospholipids. The process is typically monitored by off-line laboratory measurements of the free fatty acid content in the oil, and there is a demand for an automated on-line monitoring strategy to increase both yield and understanding of the process dynamics. This paper investigates the option of using Near-Infrared spectroscopy (NIRS) to monitor the enzymatic degumming reaction. A new method for balancing spectral noise and keeping the chemical information in the spectra obtained from a rapid changing chemical process is suggested. The effect of a varying measurement averaging window width (0 to 300 s), preprocessing method and variable selection algorithm is evaluated, aiming to obtain the most accurate and robust calibration model for prediction of the free fatty acid content (% (w/w)). The optimal Partial Least Squares (PLS) model includes eight wavelength variables, as found by rPLS (recursive PLS) calibration, and yields an RMSECV (Root Mean Square Error of Cross Validation) of 0.05% (w/w) free fatty acid using five latent variables.

Using evolutionary artificial neural networks in monitoring binary and polytomous logistic profiles

A statistical profile is a relationship between a quality characteristic (a response) and one or more explanatory variables to characterize quality of a process or a product. Monitoring profiles or checking the stability of profiles over time, has been extensively studied under the normal response variable, but it has paid a little attention to the profile with the non-normal response variable denoted by generalized linear models (GLM). Whereas, some of the potential applications of profile monitoring are cases where the response can be modelled using logistic profiles entailing binary, nominal and ordinal models. Also, most of existing control charts in this field have been developed by statistical approach and employing machine learning techniques have been rarely addressed in the related literature. Hence, to implement on-line process monitoring of logistic profiles, a novel artificial neural network (ANN) as a control chart with a heuristic training procedure is proposed in this paper. Performance of the proposed approach is investigated and compared using simulation studies in binary and polytomous models based on average run length (ARL) criterion. Simulation results revealed a good performance of the proposed approach. Nevertheless, to enhance the detection ability of the proposed approach more, the idea of combining run-rule which is a supplementary tool for making more sensitive control chart with final statistic is also implemented in this paper. Furthermore, a diagnostic method with machine learning schemes is employed to identify the shifted parameters in the profile. Results indicate the superior performance of the proposed approaches in most of the simulations. Finally, an example is used to illustrate the implementation of the proposed charting scheme.

An Adaptive Sampling Strategy for Online Monitoring and Diagnosis of High-Dimensional Streaming Data

Statistical process control techniques have been widely used for online process monitoring and diagnosis of streaming data in various applications, including manufacturing, healthcare, and environmental engineering. In some applications, the sensing system that collects online data can only provide partial information from the process due to resource constraints. In such cases, an adaptive sampling strategy is needed to decide where to collect data while maximizing the change detection capability. This article proposes an adaptive sampling strategy for online monitoring and diagnosis with partially observed data. The proposed methodology integrates two novel ideas (i) the recursive projection of the high-dimensional streaming data onto a low-dimensional subspace to capture the spatio-temporal structure of the data while performing missing data imputation; and (ii) the development of an adaptive sampling scheme, balancing exploration and exploitation, to decide where to collect data at each acquisition time. Through simulations and two case studies, the proposed framework’s performance is evaluated and compared with benchmark methods.

An Intelligent Process Fault Diagnosis System Based on Neural Networks and Andrews Plot

This paper proposes a neural network-based process fault diagnosis system with Andrews plot for information pre-processing to enhance the performance of online process fault diagnosis. By using features extracted from Andrews plot as the inputs to a neural network, as a classifier, the diagnosis speed and reliability are improved. A method for determining the important features in the Andrews function is proposed. The proposed fault diagnosis system is applied to a simulated continuous stirred tank reactor process and is compared with two conventional neural network-based fault diagnosis systems: scheme B where the monitored measurements are directly fed to a neural network after scaling and scheme C where the monitored measurements are converted to qualitative trend data before feeding to a neural network. Of all the considered faults, the proposed fault diagnosis system diagnosed the abrupt faults on average 5.45 s and 2.66 s earlier than schemes B and C, respectively and diagnosed the incipient faults on average 13.82 s and 5.09 s earlier than schemes B and C, respectively. The results reveal that Andrews plot method utilized in online process monitoring has a great potential in industrial process monitoring.

A Fully Integrated Online Platform For Real Time Monitoring Of Multiple Product Quality Attributes In Biopharmaceutical Processes For Monoclonal Antibody Therapeutics

In response to FDA's call for Quality by Design (QbD) in biopharmaceutical product development, the biopharmaceutical industry has been developing highly sensitive and specific technologies in the monitoring and controlling of product quality attributes for bioprocesses. We previously published the successful application of an off-line multi-attribute method (MAM) to monitor more than 20 critical quality attributes (CQA) with superior sensitivity for the upstream process. To further remove the hurdles of laborious process sampling and sample preparation associated with the offline method, we present here a fully integrated MAM based online platform for automated real time online process monitoring. This integrated system includes Modular Automated Sampling Technology (MAST) based aseptic sampling, multi-function Sequential Injection Analysis (SIA) sample preparation, UHPLC separation and high-resolution mass spectrometry (HRMS) analysis. Continuous automated daily monitoring of a 17-day cell culture process was successfully demonstrated for a model monoclonal antibody (mAb) molecule with similar specificity and sensitivity as we reported earlier. To the best of our knowledge, this is the first report of an end-to-end automated online MAM system, which would allow the MAM to be applied to routine bioprocess monitoring, potentially replacing multiple conventional low resolution and low sensitivity off-line methods. The online HPLC or HPLC/MS platform could be easily adapted to support other processing steps such as downstream purification with minimal software re-configuration.

Real-time synchronization with expected distribution of synchronized index for on-line monitoring of uneven multiphase batch process

Because of the inconsistent batch duration, identifying the progress of an ongoing batch highly affects the performance of the on-line monitoring. In this paper, a real-time synchronization scheme is proposed based on relative proceeding rates, newly defined as the time difference between the trajectory of an ongoing batch and the nominal trajectory. As the relative proceeding rate indicates how early the point attains its desired value compared to the nominal trajectory, the synchronized index is obtained on-line by estimating the relative proceeding rate. Compared with the conventional methods that optimize the synchronized index with local constraints, the proposed method can estimate its distribution using the relative proceeding rate without applying local constraints. It can reduce synchronization errors and increase the accuracies of the state estimation and fault detection. The utility of the proposed method is verified via a case study of industrial penicillin production with a real-time unmeasurable variable.

ANN prediction of particle flow characteristics in a drum based on synthetic acoustic signals from DEM simulations

Rotating drums are widely used in industries for particle mixing, granulation and grinding. Linking internal particle flow condition with externally measured variables is crucial to online process monitoring and control. This work proposed a modelling framework to use an artificial neural network (ANN) model for quick prediction of particle flow based on the acoustic emission (AE) signals generated from the discrete element method (DEM) simulations. In total 131 DEM simulations were conducted under different conditions (i.e., different particle size distributions and filling levels). The AE signals on the drum surface were then obtained based on the simulated particle–wall collisions. Through FFT transformation and principal component analysis (PCA), 5 principal components (PCs) were obtained and, together with power draw, fed into the ANN model to predict to the unmeasurable internal flow conditions, including filling level and the distributions of particle size and internal collision energy. The back propagation neural network was adopted in the model. After being trained with 90 datasets, the ANN model was able to predict those internal variables with reasonable accuracy (R2 > 0.95). Finally, the potentials and limitations of the model to the optimal operation of drums were discussed.

Bayesian hierarchical modeling for online process monitoring and quality control, with application to real time image data

With the increasing availability of real-time process data, the Bayesian framework is an adaptable means for dealing with complex models and data structures which can provide readily interpretable outputs for first-principles and hybrid models. Applied to process monitoring and control scenarios, Bayesian models have the capability to provide informed decisions about quantities of interest derived from the online process measurements through the use of probability as a measure of uncertainty. This work presents one such application, wherein real-time image data is used to predict the content of pharmaceutical dose units produced by an additive manufacturing process; a probabilistic quality control strategy is therefore enacted at the level of each dosage unit. Utilizing a hierarchical modeling approach, we are able to draw upon a large historical database to model sources of uncertainty across a wide range of conditions, thus yielding more reliable predictions for future data at in-sample and out-of-sample conditions.

Online monitoring and control of upstream cell culture process using 1D and 2D-LC with SegFlow interface.

The biopharmaceutical industry is transitioning from currently deployed batch-mode bioprocessing to a highly efficient and agile next-generation bioprocessing with the adaptation of continuous bioprocessing, which reduces capital investment and operational costs. Continuous bioprocessing, aligned with FDA's quality-by-design platform, is designed to develop robust processes to deliver safe and effective drugs. With the deployment of knowledge-based operations, product quality can be built into the process to achieve desired critical quality attributes (CQAs) with reduced variability. To facilitate next-generation continuous bioprocessing, it is essential to embrace a fundamental shift-in-paradigm from "quality-by-testing" to "quality-by-design," which requires the deployment of process analytical technologies (PAT). With the adaptation of PAT, a systematic approach of process and product understanding and timely process control are feasible. Deployment of PAT tools for real-time monitoring of CQAs and feedback control is critical for continuous bioprocessing. Given the current deficiency in PAT tools to support continuous bioprocessing, we have integrated Infinity 2D-LC with a post-flow-splitter in conjunction with the SegFlow autosampler to the bioreactors. With this integrated system, we have established a platform for online measurements of titer and CQAs of monoclonal antibodies as well as amino acid analysis of bioreactor cell culture.

Laser Metal Deposition On-Line Monitoring via Plasma Emission Spectroscopy and Spectral Correlation Techniques

Plasma spectroscopic techniques focused on the analysis of the plasma background radiation have been studied to enable an efficient on-line monitoring of a laser metal deposition process. The influence of different process parameters and elements, such as laser power, process speed, powder feeding rate and different powder and substrate compositions has been analyzed by means of several experimental trials. The resulting cladding patch analyzes via visual inspection and macrographs have been correlated with their associated spectroscopic monitoring signals. These studies have indicated that on-line quality monitoring of the laser metal deposition process is feasible by means of the proposed solutions, avoiding the identification and use of plasma emission lines. The latter improves the computational performance and avoids, not only the identification of each emission line, but also their specific sensitivity to certain defects. Spectral correlation techniques have also been proposed for monitoring purposes, thus enabling a more quantitative analysis.

Real-time and on-line monitoring of ethanol fermentation process by viable cell sensor and electronic nose

In this study, introduction of a viable cell sensor and electronic nose into ethanol fermentation was investigated, which could be used in real-time and on-line monitoring of the amount of living cells and product content, respectively. Compared to the conventional off-line biomass determination, the capacitance value exhibited a completely consistent trend with colony forming units, indicating that the capacitance value could reflect the living cells in the fermentation broth. On the other hand, in comparison to the results of off-line determination by high-performance liquid chromatography, the ethanol concentration measured by electronic nose presented an excellent consistency, so as to realize the on-line monitoring during the whole process. On this basis, a dynamic feeding strategy of glucose guided by the changes of living cells and ethanol content was developed. And consequently, the ethanol concentration, productivity and yield were enhanced by 15.4%, 15.9% and 9.0%, respectively. The advanced sensors adopted herein to monitor the key parameters of ethanol fermentation process could be readily extended to an industrial scale and other similar fermentation processes.

Variable-weighted FDA combined with t-SNE and multiple extreme learning machines for visual industrial process monitoring

The visualization of an operating state of industrial processes allows operators to identify and diagnose faults intuitively and quickly. The identification and diagnosis of faults are important for ensuring industrial production safety. A method that combines variable-weighted Fisher discriminant analysis (VWFDA), t-distributed stochastic neighbor embedding (t-SNE), and multiple extreme learning machines (ELMs) is proposed for visual process monitoring. First, the VWFDA weighs variables on the basis of their contribution to the fault, thereby amplifying the fault information. The VWFDA is used to extract feature vectors from industrial data, and normal state and various fault states can be separated from each other in the space formed by these feature vectors. Second, t-SNE is used to visualize these feature vectors. Third, given that t-SNE lacks a transformation matrix during dimension reduction, one ELM is used for each class data of t-SNE to obtain the mapping relation from its input data to its mapping points. Finally, the VWFDA and multiple trained ELMs are combined for online process monitoring. The performance of the proposed approach is compared with that of FDA–t-SNE and other methods on the basis of the Tennessee Eastman process, thereby confirming that the proposed approach is advantageous for visual industrial process monitoring.

In situ monitoring and penetration prediction of plasma arc welding based on welder intelligence-enhanced deep random forest fusion

Online process monitoring and quality control has been a long-standing challenge for variable polarity plasma arc welding (VPPAW) due to the inherent instability and fluctuation of the keyhole molten pool. This work developed an innovative welder intelligence-enhanced deep random forest fusion (WI-DRFF) approach, aiming to describe the dynamics of front-side molten pool and accurately predict the weld penetration. Based on the human welder’s prior knowledge, we firstly proposed an image processing algorithm to extract the low-level handcrafted features, which could quantitatively describe the geometrical appearance of the keyhole. Afterwards, we constructed a convolutional neural network (CNN) to learn the high-level discriminative features of weld pool and interpret the physical characteristics of the deep features with visualization. Finally, we incorporated the handcrafted keyhole features and deep features to concatenate a multi-level feature vector for predicting the weld penetration based on random forest (RF) classifier. Extensive experiments demonstrate that our proposed approach yields a remarkable classification performance comparing with state-of-the-art machine learning algorithms even with limited training data. This approach is a new paradigm in the digitization and intelligence of welding process and can be exploited to provide a feedback in an adaptive quality control system.

On-line monitoring of penetration state in laser-arc hybrid welding based on keyhole and arc features

Incomplete penetration is a typical defect in laser-arc hybrid welding. On-line monitoring of welding process is an important method to assess welding quality. In laser-arc hybrid welding, there is a strong correlation between keyhole, arc and incomplete penetration. Therefore, an on-line monitoring method of penetration state based on keyhole and arc features is proposed in this paper. In the proposed method, the images of keyhole and arc in laser-arc hybrid welding are captured by high-speed camera, and then the keyhole and arc features are extracted by image processing algorithm. Finally, the features are used as input to classify the penetration state using SVM model. The results show that the SVM model based on keyhole and arc features can accurately identify the penetration state.

Estimating uncertainties and parameters for fundamental models used in online monitoring and control

AbstractMany model‐based online process monitoring and control applications rely on state estimation techniques that use noisy process data to update states, thereby ensuring that imperfect model predictions are consistent with process behaviour. Techniques for tuning state estimators are reviewed, and their effectiveness and limitations are summarized in this article. A new simultaneous parameter and estimator tuning (SPET) methodology is proposed, one in which parameter estimation techniques for stochastic differential equations (SDEs) are used to simultaneously estimate measurement‐error covariances and model‐error covariances along with the model parameters. The resulting information is then used to compute state‐estimator tuning information. This study shows how SPET can be used, along with old dynamic process data, to obtain reliable tuning information. The proposed methodology is tested using a nonlinear two‐state continuous stirred‐tank reactor (CSTR) model with simulated data. Comparisons are made with a more‐conventional approach that uses WLS to estimate fixed model parameters and autocovariance least squares (ALS) to estimate extended Kalman filter (EKF) tuning factors. The main benefit of the proposed approach is that fixed model parameters and tuning factors for the EKF are estimated simultaneously, resulting in significant improvements to state estimates and online model predictions.

Robotic automated fiber placement of carbon fiber towpregs

ABSTRACT Robotic Automated Fiber Placement (R-AFP) technology was developed to manufacture composite laminates by placing carbon fiber thermoset towpregs, obtained by impregnating 12 K high-strength grade carbon fibers in an epoxy resin system. In order to avoid placement induced defects, a thermographic scanning technique was implemented for on-line quality monitoring of the R-AFP process. The thermal analysis proved to be an effective and quick approach for the real-time detection of deposition defects generated during. The R-AFP process was performed by applying a constant pressure through the compaction roller; different pressure values were investigated. The effect of the resin weight fraction and compaction pressure on the mechanical properties applied by the deposition head during R-AFP of cross-ply laminates was studied. The reduction of the tensile strength and the increase of the elastic modulus with decreasing pressure of the compaction roller was observed. Furthermore, the values of ultimate tensile strength and elastic modulus decrease as the resin content increases. Finally, the three-dimensional topography of surface fracture of tensile samples was investigated by means of the optical and scanning electron microscopy.

An integrated manifold learning approach for high-dimensional data feature extractions and its applications to online process monitoring of additive manufacturing

As an effective dimension reduction and feature extraction technique, manifold learning has been successfully applied to high-dimensional data analysis. With the rapid development of sensor technology, a large amount of high-dimensional data such as image streams can be easily available. Thus, a promising application of manifold learning is in the field of sensor signal analysis, particular for the applications of online process monitoring and control using high-dimensional data. The objective of this study is to develop a manifold learning-based feature extraction method for process monitoring of Additive Manufacturing (AM) using online sensor data. Due to the non-parametric nature of most existing manifold learning methods, their performance in terms of computational efficiency, as well as noise resistance has yet to be improved. To address this issue, this study proposes an integrated manifold learning approach termed multi-kernel metric learning embedded isometric feature mapping (MKML-ISOMAP) for dimension reduction and feature extraction of online high-dimensional sensor data such as images. Based on the extracted features with the utilization of supervised classification and regression methods, an online process monitoring methodology for AM is implemented to identify the actual process quality status. In the numerical simulation and real-world case studies, the proposed method demonstrates excellent performance in both prediction accuracy and computational efficiency.

A 3D Deep Learning Model for Rapid Prediction of Structural Dynamics of Workpieces During Machining

Chatter stability in machining flexible parts depends directly on the structural dynamics of the workpiece. This paper proposes a novel two-stage framework that combines finite element (FE) and data-driven deep learning techniques to rapidly predict the varying dynamics of workpieces during machining. An automated framework is developed to create a large training dataset of CAD models with gradually-changing geometries. A deep 3D Convolutional Neural Network (3D-CNN) is developed to “learn” the variations in dynamic parameters as a function of geometry. The current model has been successfully implemented for prediction of natural frequencies of workpieces during turning operations. The proposed framework can be used as a computationally efficient tool in online process monitoring and automated correction applications.