Product Quality Prediction Research Articles

Adaptive Online Continual Learning for In-Situ Quality Prediction in Manufacturing Processes

Abstract Manufacturing processes undergo continuous changes in order to meet various requirements, such as process/product changes and variations in tool/workpiece conditions, leading to mixed, heterogenous, or anomalous data. As a result, a quality prediction model trained from previous data may not perform well when new tasks emerge. In order to achieve in-time and accurate product quality prediction, it is crucial to develop a predictive method that adapt to variations in the manufacturing system, capable of learning from new tasks without forgetting previous ones and detecting unknown tasks. This study proposes a deep learning method integrated with continual learning for in-situ quality prediction that is capable of learning from new tasks without forgetting previous ones. To demonstrate this idea, deep CNNs are designed to analyze in-process sensor data, which consist of shared layers to capture the common underlying features across all tasks, and task-specific layers that capture specific characteristics of each individual task. In order to identify the task to which incoming product belongs, a task prediction approach based on task relevancy using filter subspace distance is proposed. When new data come in, the model first identifies the task, followed by predicting the quality of the current product. The proposed method is demonstrated on two case studies, including quality prediction of the workpiece using acoustic emissions during the laser-induced plasma micro-machining process, and quality prediction of the product through thermal images during the hot stamping process.

Effect of process parameters on the mechanical properties of wires produced from A356 aluminum alloy chips by Continuous Friction Stir Extrusion: Experiments and numerical simulation

Recycling of metals is becoming crucial from an economic and environmental point of view. The solid-state recycling process Continuous Friction Stir Extrusion was used to produce wires out of A356-T6 chips. The mechanical properties of the produced wires were explored by varying the main process parameters. Characterization involved Vickers hardness tests, tensile tests, grain size measurements, and fracture surface analysis. It has been found that it is possible to achieve 77 % of the Ultimate Tensile Strength (UTS) and 92 % of Vickers hardness with respect to the as-fabricated A356 alloy. The average grain size increases with the tool rotational with values ranging from about 9 µm to about 11 µm. A 3D dedicated numerical model was used to predict the distributions and histories of primary field variables, and to calculate the Piwnik-Plata parameter, fostering a more in-depth understanding of the process mechanics. This allows for the precise prediction of unacceptable product quality of the bonding when the Plata and Piwnik parameters are low. Predicted temperature close to the rotating tool should reach 400 °C while the cochlea temperature should be below 100 °C for sound wires production thus avoiding early chip bonding and process failure.

Multiobjective Ensemble Learning With Multiscale Data for Product Quality Prediction in Iron and Steel Industry

Multiobjective Ensemble Learning With Multiscale Data for Product Quality Prediction in Iron and Steel Industry

Data‐driven quality prediction in injection molding: An autoencoder and machine learning approach

AbstractIn the injection molding process, the pressure within the mold cavity is crucial to the quality of the final product. Due to the inability to directly observe the process, sensor technology is required to acquire data. Traditionally, experts interpret and encode pressure curves, but this method has limitations. This study proposes an innovative pressure curve encoding technique to overcome these limitations and achieve automation to obtain more comprehensive pressure information. The study employs mold flow analysis software and autoencoders to capture and encode pressure data, classifying pressure curves into global pressure and local pressure values. Subsequently, a multilayer perceptron (MLP) neural network is used for machine learning to predict multiple qualities. Results indicate that local pressure features perform better in predicting multiple‐quality targets than global pressure features, exhibiting smaller prediction ranges and higher prediction stability. Although domain knowledge‐based indicator features slightly outperform in terms of predictive capability, the low error results of the local pressure feature method validate the effectiveness of the autoencoder approach, demonstrating its potential for digital information extraction and practical quality prediction in the injection molding process.Highlights Develops a product quality prediction system for efficient injection molding. Autoencoders extract key features from pressure data without domain knowledge. ML models predict quality indicators, optimizing injection molding processes. Compares pressure features, showing data‐driven methods' prediction accuracy.

Enhancing SMEs digital transformation through machine learning: A framework for adaptive quality prediction

As smart manufacturing expands, businesses see the importance of digital transformation, especially for small and medium-sized enterprises (SMEs). Unlike larger companies, SMEs face greater challenges when undergoing digital transformation due to technological questions. However, recent advancements in high-performance computing and reduced hardware costs have made deep learning-based digital transformation more financially feasible for SMEs. While previous research utilized machine learning for product quality prediction, there remains a lack of comprehensive in adaptive quality prediction specifically designed for SMEs. This study presents a systematic framework utilizing various machine learning methods and validates research cases using CRISP-DM (Cross-Industry Standard Process for Data Mining). The first step involves applying XGBoost(eXtreme Gradient Boosting)for feature selection, the second step utilizes GRU for parameter prediction. Finally, in the third step, SVM (Support Vector Machine) is employed for quality classification. The integrated framework achieves high accuracy, with R2reaching 90 % for predicted parameters and nearly 95 % for classification indicators. Moreover, this research addresses the research gap in quality prediction and adaptability and provide an effective digital transformation solution for SMEs without substantial investment. The proposed research framework can be applied SMEs of other domains, such as the machining and traditional manufacturing industry.

A Multiphase Dual Attention-Based LSTM Neural Network for Industrial Product Quality Prediction

A Multiphase Dual Attention-Based LSTM Neural Network for Industrial Product Quality Prediction

Implementing multiblock techniques in a full-scale plant scenario: On-line prediction of quality parameters in a continuous process for different acrylonitrile butadiene styrene (ABS) products

BackgroundThe study explores the challenges of handling multiblock data of different natures (process and NIR sensors) for on-line quality prediction in a full-scale plant scenario, namely a plant operating in continuous on an industrial scale and producing different grade Acrylonitrile Butadiene Styrene (ABS) products. This environment is an ideal scenario to evaluate the use of multiblock data analysis methods, which can enhance data interpretation, visualization, and predictive performances. In particular, a novel multiblock extension of Locally Weighted PLS has been proposed by the authors, namely Locally Weighted Multiblock Partial Least Squares (LW-MB-PLS). Response-Oriented Sequential Alternation (ROSA) has also been employed to evaluate the diverse block relevance for the prediction of two quality parameters associated with the polymer. Data are split in blocks both according to sensor type and different plant sections, and different models have been built by incremental addition of data blocks to evaluate if early estimation of product quality is feasible. ResultsROSA method showed promising predictive performance for both quality parameters, highlighting the most influential plant sections through the selection of data blocks. The results suggested that both early and late-stage sensors play crucial roles in predicting product quality. A reasonable estimation of quality parameters before production completion has been achieved. On the other hand, the proposed LW-MB-PLS, while comparable in predictive performances, allowed reducing systematic prediction errors for specific products. SignificanceThis study contributes valuable insights for continuous production processes, aiding plant operators and paving the way for advancements in online quality prediction and control. Furthermore, it is implemented as a locally weighted extension of MB-PLS.

Small-batch product quality prediction using a novel discrete Choquet fuzzy grey model with complex interaction information

With manufacturing focused on multi-variety and small-batch intelligent production, the challenge is to guarantee and ensure good product quality. This paper proposes a novel small-batch product quality prediction approach considering complex interaction information. In this paper, first, to handle the complex nonlinear correlations among the manufacturing process variables, a kernel principal components analysis is used to extract the major features from the high-dimensional data. Second, we combine the discrete multivariate grey model and generalized discrete Choquet fuzzy integral to propose a discrete Choquet fuzzy grey model so as to capture the complex interaction information among the process variables. Next, an adjoint sensitivity analysis is applied to characterize the changes in the process variables on the predicted product quality index. We conduct an experiment on six semiconductor products made in China to validate the proposed model against a library of nine other methods. Our model yields an average reduction of 11.02% on MAPE, 20.47% on MSE, 13.56% on STD, and an average increase of 46.37% on EVS for six types of products. Our results inform that the proposed model is suitable for predicting manufacturing product quality when the process variables interact significantly, and we can prioritize the process variables for remedial action accordingly.

Developing a Model for Product Quality Prediction and Improvement Using the Decision Tree Algorithm and Data Envelopment Analysis – a Case Study: Manufacturers of the Tiba Anti-roll Bar in Iran

Developing a Model for Product Quality Prediction and Improvement Using the Decision Tree Algorithm and Data Envelopment Analysis – a Case Study: Manufacturers of the Tiba Anti-roll Bar in Iran

Uncertainty Quantification of Data-driven Quality Prediction Model For Realizing the Active Sampling Inspection of Mechanical Properties in Steel Production

Pre-production quality defect inspection is a crucial step in industrial manufacturing, and many traditional inspection strategies suffer from inefficiency issues. This is especially true for tasks such as mechanical performance testing of steel products, which involve time-consuming processes like offline sampling, specimen preparation, and testing. The inspection volume significantly impacts the production cycle, inventory, yield, and labor costs. Constructing a data-driven model for predicting product quality and implementing proactive sampling inspection based on the prediction results is an appealing solution. However, the prediction uncertainty of data-driven models poses a challenging problem that needs to be addressed. This paper proposes an active quality inspection approach for steel products based on the uncertainty quantification in the predictive model for mechanical performance. The objective is to reduce both the sampling frequency and the omission rate on the production site. First, an ensemble model based on improved lower and upper bound estimation is established for interval prediction of mechanical performance. The uncertainty of the specific value prediction model is quantitatively estimated using interval probability distributions. Then, a predictive model for the mechanical performance failure probability is built based on the prediction interval size and probability distribution. By determining an appropriate probability threshold, the trade-off between prediction accuracy and defect detection accuracy (recall rate) is balanced, enabling the establishment of an active sampling strategy. Finally, this functionality is integrated into the manufacturing execution system of a steel factory, realizing a mechanical performance inspection approach based on proactive sampling. The proposed approach is validated using real production datasets. When the probability threshold is set to 30%, the prediction accuracy and recall rate for failure mechanical performance samples are 75% and 100%, respectively. Meanwhile, the sampling rate is only 5.33%, while controlling the risk of omission. This represents a 50% reduction in sampling rate compared to the inspection rules commonly used in actual production. The overall efficiency of product quality inspection is improved, and inspection costs are reduced.

Data-driven quasi-convex method for hit rate optimization of process product quality in digital twin

Hit rate is an important quantitative criterion for the process product quality prediction of the integrated industrial processes. The hit rate indicates the percentage of product quantities accepted by the downstream process within the controlled range of the product quality. The optimization model of the hit rate criterion is a non-convex intractable problem. In order to improve the hit rate of the predicted product quality, we define a hit rate optimization problem, and propose a data-driven quasi-convex approach, which converts the original problem into a set of convex feasible problems and achieves the optimal hit rate. The proposed approach combines factorial hidden Markov models, multitask elastic net and quasi-convex optimization. In order to illustrate the advantages of the proposed approach, a Monte Carlo simulation experiment is designed to verify the convex optimization property. Another experiment is carried out on two actual steel production datasets for the temperature prediction in molten iron dispatch. The results confirm that the proposed approach not only exhibits superior performance with the controlled hit rate, but also improves the hit rate by at least 41.11 % and 31.01 %, respectively, compared with the classical models on two real datasets.

Data-Driven Modeling and Operation Optimization With Inherent Feature Extraction for Complex Industrial Processes

In response to the tenets of Industry 4.0, operation optimization in industrial processes has become a significant research topic. However, the uncertainties prevailing in the process pose challenges to production operations, especially the feedstock properties. In this work, the operation optimization study is performed on a distillation unit (DU), a typical plant in the industrial process. To enhance production performance, a modeling and operation optimization strategy based on feedstock property and production features is presented. One of the difficulties is how to uncover features from high-dimensional and imperfect data, where imperfect data refers to product quality data that is unavailable online. In the strategy, we inject the inherent characteristic of the process into the data-driven method to extract the feedstock property in a data-based and knowledge-oriented manner. Further, optimal feature representation and process modeling can be achieved by customizing the network structure. The operation optimization problem is formulated to adjust the top temperature of the distillation column (TTDC) to achieve satisfactory production under varying feedstock properties. Experimental results illustrate that the process model based on feedstock property and production features (PM-FP-PF) can better fit the physical process mechanism even based on incomplete information in industrial data. Industrial experiments have shown the proposed strategy has advanced generalization ability to the different feedstock properties. The proposed operation optimization strategy (OOS) improves the product qualification rate and has broad application prospects in industrial processes with similar features. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Industrial processes suffer from a variety of disturbances that interrupt the smooth operation of the system, such as varying feedstock properties. How to deal with them is the key to improve the product qualification rate. In this work, we propose a data-driven modeling and operation optimization framework to improve product quality under varying feedstock properties. The dynamic variation characteristics of the feedstock properties can be obtained by analyzing the physical properties of the production unit. This is used as a basis for representing and extracting feedstock properties in a data-driven way. Further, a process model based on feedstock property and production features (PM-FP-PF) is built to predict product quality. An operation optimization strategy with production capacity consideration is established. It can determine the optimal operation action required by the current system, mitigating the uncertainty of feedstock property. This operation optimization system has been applied to a distillation unit in the hydrofining process. The application results show that the process model achieves satisfactory estimation accuracy, and the operation optimization strategy has improved production performance.

Digital twin enhanced quality prediction method of powder compaction process

During the powder compaction process, process parameters are required for product quality prediction. However, the inadequacy of compaction data leads to difficulties in constructing models for quality prediction. Meanwhile, the existing data generation methods can only generate required data partially, and fail to generate data for extreme operating conditions and difficult-to-measure quality parameters. To address this issue, a digital twin (DT) enhanced quality prediction method for powder compaction process is presented in this paper. First, a DT model of the powder compaction process with multiple dimensions is constructed and validated. Then, to solve the data inadequacy problem, data of process parameters are generated through an orthogonal experimental design, and are imported into the DT model to generate quality parameters, so as to obtain the virtual data. Finally, the quality prediction for the powder compaction process is achieved by the generative adversarial network-deep neural network (GAN-DNN) method. The effectiveness of the generated virtual data and the GAN-DNN method is verified through experimental comparison. On top of point-to-point validation, a quality prediction system applied in a powder compaction line is developed and implemented to demonstrate the end-to-end practicability of the proposed method.

Quality prediction of whole-grain rice noodles using backpropagation artificial neural network.

Whole-grain rice noodles are a kind of healthy food with rich nutritional value, and their product quality has a notable impact on consumer acceptability. The quality evaluation model is of great significance to the optimization of product quality. However, there are few methods that can establish a product quality prediction model with multiple preparation conditions as inputs and various quality evaluation indexes as outputs. In this study, an artificial neural network (ANN) model based on a backpropagation (BP) algorithm was used to predict the comprehensive quality changes of whole-grain rice noodles under different preparation conditions, which provided a new way to improve the quality of extrusion rice products. The results showed that the BP-ANN using the Levenberg-Marquardt algorithm and the optimal topology (4-11-8) gave the best performance. The correlation coefficients (R2) for the training, validation, testing, and global data sets of the BP neural network were 0.927, 0.873, 0.817, and 0.903, respectively. In the validation test, the percentage error in the quality prediction of whole-grain rice noodles was within 10%, indicating that the BP-ANN could accurately predict the quality of whole-grain rice noodles prepared under different conditions. This study showed that the quality prediction model of whole-grain rice noodles based on the BP-ANN algorithm was effective, and suitable for predicting the quality of whole-grain rice noodles prepared under different conditions. © 2024 Society of Chemical Industry.

An innovative hybrid modeling approach for simultaneous prediction of cell culture process dynamics and product quality.

The use of hybrid models is extensively described in the literature to predict the process evolution in cell cultures. These models combine mechanistic and machine learning methods, allowing the prediction of complex process behavior, in the presence of many process variables, without the need to collect a large amount of data. Hybrid models cannot be directly used to predict final product critical quality attributes, or CQAs, because they are usually measured only at the end of the process, and more mechanistic knowledge is needed for many classes of CQAs. The historical models can instead predict the CQAs better; however, they cannot directly relate manipulated process parameters to final CQAs, as they require knowledge of the process evolution. In this work, we propose an innovative modeling approach based on combining a hybrid propagation model with a historical data-driven model, that is, the combined hybrid model, for simultaneous prediction of full process dynamics and CQAs. The performance of the combined hybrid model was evaluated on an industrial dataset and compared to classical black-box models, which directly relate manipulated process parameters to CQAs. The proposed combined hybrid model outperforms the black-box model by 33% on average in predicting the CQAs while requiring only around half of the data for model training to match performance. Thus, in terms of model accuracy and experimental costs, the combined hybrid model in this study provides a promising platform for process optimization applications.

From vineyard to table: Uncovering wine quality for sales management through machine learning

The literature currently offers limited guidance for retailers on how to use analytics to decipher the relationship between product attributes and quality ratings. Addressing this gap, our study introduces an advanced ensemble learning approach to develop a nuanced framework for assessing product quality. We validated the effectiveness of our framework with a dataset comprising 1,599 red wine samples from Portugal’s Minho region. Our findings show that this model surpasses previous ones in accurately predicting product quality, presenting retailers with a sophisticated tool to transform product data into actionable insights for sales management. Furthermore, our approach yields significant benefits for researchers by identifying latent attributes in extensive data collections, which can inform a deeper understanding of consumer preferences and guide the strategic planning of marketing promotions.

Cloud-Based Machine Learning Methods for Parameter Prediction in Textile Manufacturing.

In traditional textile manufacturing, downstream manufacturers use raw materials, such as Nylon and cotton yarns, to produce textile products. The manufacturing process involves warping, sizing, beaming, weaving, and inspection. Staff members typically use a trial-and-error approach to adjust the appropriate production parameters in the manufacturing process, which can be time consuming and a waste of resources. To enhance the efficiency and effectiveness of textile manufacturing economically, this study proposes a query-based learning method in regression analytics using existing manufacturing data. Query-based learning allows the model training to evolve its decision-making process through dynamic interactions with its solution space. In this study, predefined target parameters of quality factors were first used to validate the training results and create new training patterns. These new patterns were then imported into the solution space of the training model. In predicting product quality, the results show that the proposed query-based regression algorithm has a mean squared error of 0.0153, which is better than those of the original regression-related methods (Avg. mean squared error = 0.020). The trained model was deployed as an application programing interface (API) for cloud-based analytics and an extensive auto-notification service.

Product Quality Prediction Based on RBF Optimized by Firefly Algorithm

With the development of information technology, a large number of product quality data in the entire manufacturing process is accumulated, but it is not explored and used effectively. The traditional product quality prediction models have many disadvantages, such as high complexity and low accuracy. To overcome the above problems, we propose an optimized data equalization method to pre-process dataset and design a simple but effective product quality prediction model: radial basis function model optimized by the firefly algorithm with Levy flight mechanism (RBFFALM). First, the new data equalization method is introduced to pre-process the dataset, which reduces the dimension of the data, removes redundant features, and improves the data distribution. Then the RBFFALFM is used to predict product quality. Comprehensive expe riments conducted on real-world product quality datasets validate that the new model RBFFALFM combining with the new data pre-processing method outperforms other previous me thods on predicting product quality.

Edge computing-based proactive control method for industrial product manufacturing quality prediction

With the emergence of intelligent manufacturing, new-generation information technologies such as big data and artificial intelligence are rapidly integrating with the manufacturing industry. One of the primary applications is to assist manufacturing plants in predicting product quality. Traditional predictive models primarily focus on establishing high-precision classification or regression models, with less emphasis on imbalanced data. This is a specific but common scenario in practical industrial environments concerning quality prediction. A SMOTE-XGboost quality prediction active control method based on joint optimization hyperparameters is proposed to address the problem of imbalanced data classification in product quality prediction. In addition, edge computing technology is introduced to address issues in industrial manufacturing, such as the large bandwidth load and resource limitations associated with traditional cloud computing models. Finally, the practicality and effectiveness of the proposed method are validated through a case study of the brake disc production line. Experimental results indicate that the proposed method outperforms other classification methods in brake disc quality prediction.

Sustainability in Semiconductor Production via Interpretable and Reliable Predictions

Sustainability in production stands out as one of the foremost achievements facilitated by Industry 4.0. With the continuous monitoring of production lines, it is now possible to detect quality deviations at their earliest occurrence and reduce the scrap production rate. This paper studies the sustainability in context of a partially monitored multistage semiconductor production line with multiple test units. Our goal is to provide a foundation for interpretable and reliable Product Quality Prediction (PQP) in industrial use cases with noisy quality check results and missing process information. To that end, we examine how the accumulated process information from different steps of the production line impacts the accuracy of the PQP models used later as base learners. Furthermore, we highlight the drawbacks of conventional model stacking on the accuracy due to base learner’s prediction quantization. We propose instead an interpretable stacked model (SM) from base learners’ predicted probability values, and demonstrate how it increases the accuracy and reliability of the predictions. To improve the results even further, we analyze different calibration methods both for base learners and the SM. Our final model leads to a 19.49% reduction in the binary estimated calibration error compared to the conventional models, and thus allows for increased PQP reliability.