Online monitoring and control of a cyber-physical manufacturing process under uncertainty.

Recent technological advancements in computing, sensing and communication have led to the development of cyber-physical manufacturing processes, where a computing subsystem monitors the manufacturing process performance in real-time by analyzing sensor data and implements the necessary control to improve the product quality. This paper develops a predictive control framework where control actions are implemented after predicting the state of the manufacturing process or product quality at a future time using process models. In a cyber-physical manufacturing process, the product quality predictions may be affected by uncertainty sources from the computing subsystem (resource and communication uncertainty), manufacturing process (input uncertainty, process variability and modeling errors), and sensors (measurement uncertainty). In addition, due to the continuous interactions between the computing subsystem and the manufacturing process, these uncertainty sources may aggregate and compound over time. In some cases, some process parameters needed for model predictions may not be precisely known and may need to be derived from real time sensor data. This paper develops a dynamic Bayesian network approach, which enables the aggregation of multiple uncertainty sources, parameter estimation and robust prediction for online control. As the number of process parameters increase, their estimation using sensor data in real-time can be computationally expensive. To facilitate real-time analysis, variance-based global sensitivity analysis is used for dimension reduction. The proposed methodology of online monitoring and control under uncertainty, and dimension reduction, are illustrated for a cyber-physical turning process.

Fast Surrogate Modeling using Dimensionality Reduction in Model Inputs and Field Output: Application to Additive Manufacturing.

A novel approach to surrogate modeling motivated by recent advancements in parameter dimension reduction is proposed. Specifically, the approach aims to speed-up surrogate modeling for mapping multiple input variables to a field quantity of interest. Computational efficiency is accomplished by first identifying principal components (PC) and corresponding features in the output field data. A map from inputs to each feature is considered, and the active subspace (AS) methodology is used to capture their relationship in a low-dimensional subspace in the input domain. Thus, the PCAS method accomplishes dimension reduction in the input as well as the output. The method is demonstrated on a realistic problem pertaining to variability in residual stress in an additively manufactured component due to the stochastic nature of the process variables and material properties. The resulting surrogate model is exploited for uncertainty propagation, and identification of stress hotspots in the part. Additionally, the surrogate model is used for global sensitivity analysis to quantify relative contributions of the uncertain inputs to stress variability. Our findings based on the considered application are indicative of enormous potential for computational gains in such analyses, especially in generating training data, and enabling advancements in control and optimization of additive manufacturing processes.

Optimization of fused filament fabrication process parameters under uncertainty to maximize part geometry accuracy.

This work presents a novel process design optimization framework for additive manufacturing (AM) by integrating physics-informed computational simulation models with experimental observations. The proposed framework is implemented to optimize the process parameters such as extrusion temperature, extrusion velocity, and layer thickness in the fused filament fabrication (FFF) AM process, in order to reduce the variability in the geometry of the manufactured part. A coupled thermo-mechanical model is first developed to simulate the FFF process. The temperature history obtained from the heat transfer analysis is then used as input for the mechanical deformation analysis to predict the dimensional inaccuracy of the additively manufactured part. The simulation model is then corrected based on experimental observations through Bayesian calibration of the model discrepancy to make it more accurately represent the actual manufacturing process. Based on the corrected prediction model, a robustness-based design optimization problem is formulated to optimize the process parameters, while accounting for multiple sources of uncertainty in the manufacturing process, process models, and measurements. Physical experiments are conducted to verify the effectiveness of the proposed optimization framework.

Complexity and entropy representation for machine component diagnostics.

The Complexity-entropy causality plane (CECP) is a parsimonious representation space for time series. It has only two dimensions: normalized permutation entropy ([Formula: see text]) and Jensen-Shannon complexity ([Formula: see text]) of a time series. This two-dimensional representation allows for detection of slow or rapid drifts in the condition of mechanical components monitored through sensor measurements. The CECP representation can be used for both predictive analytics and visual monitoring of changes in component condition. This method requires minimal pre-processing of raw signals. Furthermore, it is insensitive to noise, stationarity, and trends. These desirable properties make CECP a good candidate for machine condition monitoring and fault diagnostics. In this work we study the effectiveness of CECP on three rotary component condition assessment applications. We use CECP representation of vibration signals to differentiate various machine component health conditions for rotary machine components, namely roller bearing and gears. The results confirm that the CECP representation is able to detect, with high accuracy, changes in underlying dynamics of machine component degradation states. From class separability perspective, the CECP representation is able to generate linearly separable classes for the classification of different fault states. This classification performance improves with increasing signal length. For signal length of 16,384 data points, the fault classification accuracy varies from 90% to 100% for bearing applications, and from 85% to 100% for gear applications. We observed that the optimum parameter for CECP representatino depends on the application. For bearing applications we found that embedding dimension D = 4, 5, 6, and embedding delay τ = 1, 2, 3 are suitable for good fault classification. For gear applications we find that embedding dimension D = 4, 5, and embedding delay τ = 1, 5 are suitable for fault classification.

A Standardized PMML Format for Representing Convolutional Neural Networks with Application to Defect Detection.

Convolutional neural networks are becoming a popular tool for image processing in the engineering and manufacturing sectors. However, managing the storage and distribution of trained models is still a difficult task, partially due to the lack of standardized methods for deep neural network representation. Additionally, the interoperability between different machine learning frameworks remains poor. This paper seeks to address this issue by proposing a standardized format for convolutional neural networks, based on the Predictive Model Markup Language (PMML). A new standardized schema is proposed to represent a range of convolutional neural networks, including classification, regression and semantic segmentation systems. To demonstrate the practical application of this standard, a semantic segmentation model, which is trained to detect casting defects in Xray images, is represented in the proposed PMML format. A high-performance scoring engine is developed to evaluate images and videos against the PMML model. The utility of the proposed format and the scoring engine is evaluated by benchmarking the performance of the defect detection models on a range of different computational platforms.

Detection and Segmentation of Manufacturing Defects with Convolutional Neural Networks and Transfer Learning.

Quality control is a fundamental component of many manufacturing processes, especially those involving casting or welding. However, manual quality control procedures are often time-consuming and error-prone. In order to meet the growing demand for high-quality products, the use of intelligent visual inspection systems is becoming essential in production lines. Recently, Convolutional Neural Networks (CNNs) have shown outstanding performance in both image classification and localization tasks. In this article, a system is proposed for the identification of casting defects in X-ray images, based on the Mask Region-based CNN architecture. The proposed defect detection system simultaneously performs defect detection and segmentation on input images, making it suitable for a range of defect detection tasks. It is shown that training the network to simultaneously perform defect detection and defect instance segmentation, results in a higher defect detection accuracy than training on defect detection alone. Transfer learning is leveraged to reduce the training data demands and increase the prediction accuracy of the trained model. More specifically, the model is first trained with two large openly-available image datasets before finetuning on a relatively small metal casting X-ray dataset. The accuracy of the trained model exceeds state-of-the art performance on the GRIMA database of X-ray images (GDXray) Castings dataset and is fast enough to be used in a production setting. The system also performs well on the GDXray Welds dataset. A number of in-depth studies are conducted to explore how transfer learning, multi-task learning, and multi-class learning influence the performance of the trained system.

Predictive Model Markup Language (PMML) Representation of Bayesian Networks: An Application in Manufacturing.

Bayesian networks (BNs) represent a promising approach for the aggregation of multiple uncertainty sources in manufacturing networks and other engineering systems for the purposes of uncertainty quantification, risk analysis, and quality control. A standardized representation for BN models will aid in their communication and exchange across the web. This paper presents an extension to the Predictive Model Markup Language (PMML) standard, for the representation of a BN, which may consist of discrete variables, continuous variables, or their combination. The PMML standard is based on Extensible Markup Language (XML) and used for the representation of analytical models. The BN PMML representation is available in PMML v4.3 released by the Data Mining Group. We demonstrate the conversion of analytical models into the BN PMML representation, and the PMML representation of such models into analytical models, through a Python parser. The BNs obtained after parsing PMML representation can then be used to perform Bayesian inference. Finally, we illustrate the developed BN PMML schema for a welding process.

A Journey in Standard Development: The Core Manufacturing Simulation Data (CMSD) Information Model.

This report documents a journey "from research to an approved standard" of a NIST-led standard development activity. That standard, Core Manufacturing Simulation Data (CMSD) information model, provides neutral structures for the efficient exchange of manufacturing data in a simulation environment. The model was standardized under the auspices of the international Simulation Interoperability Standards Organization (SISO). NIST started the research in 2001 and initiated the standardization effort in 2004. The CMSD standard was published in two SISO Products. In the first Product, the information model was defined in the Unified Modeling Language (UML) and published in 2010 as SISO-STD-008-2010. In the second Product, the information model was defined in Extensible Markup Language (XML) and published in 2013 as SISO-STD-008-01-2012. Both SISO-STD-008-2010 and SISO-STD-008-01-2012 are intended to be used together.