Data-driven Process Control Research Articles

Developing deep learning surrogate models for digital twins in mineral processing – A case study on data-driven multivariate multistep forecasting

The escalating demand for environmental and social sustainability underscores the critical need for large industries such as mining and metallurgy to function optimally. Achieving optimal operation of a mineral processing plant requires timely access to information and rapid implementation of corrective actions. Essential information, such as current and future production, plays a pivotal role in enhancing process efficiency by facilitating adjustments in key aspects, including feed, control and operating variables. This study delves into the application of deep learning algorithms to forecast the gold concentrate grade of a flotation process half an hour in advance, a valuable tool for short-term decision-making, e.g., with regard to optimising the operation and process control. Multivariate time series data from a real gold processing plant was acquired through the process instrumentation and control system and augmented with simulated data at 5-minute intervals over 236 days. Surrogate models based on LSTM, GRU and CNN architectures were developed and analysed for predicting the gold grade one to six steps ahead. Evaluation of the testing dataset indicates that the CNN exhibits superior performance, achieving an average error reduction of 36 %, 45 % and 35 % for the performance metrics RMSE, MAE and MAPE respectively compared to the baseline model. These findings reveal the high potential of deep learning surrogate methods for accurately forecasting process outputs, enabling effective data-driven process control and optimisation strategies.

Quantitative comparison of reinforcement learning and data-driven model predictive control for chemical and biological processes

As manufacturing processes transition towards digitalization, data-driven process control is emerging as a key area of interest in future artificial intelligence technology. A crucial aspect in implementing data-driven process control is “What should we learn from the data?”. In general, the data-driven control method can be categorized into two main approaches: Learning the model and learning the value. To assist in selecting the more suitable approach, this paper applies six different control methods, with three falling under each approach, to three distinct manufacturing process systems. The simulation results indicate that the model-learning approaches display higher data efficiency and exhibit lower variance in total cost. These methods prove to be particularly advantageous for addressing the regulation problems. Conversely, value-learning approaches show competitive potential in closed-loop identification and in managing economic cost problems. The remaining challenges associated with each technique are discussed, along with practical considerations for their implementation.

Data-Driven Process Analysis for Iron Foundries with Automatic Sand Molding Process

This paper proposes a methodological framework to develop a data-driven process control using pure industrial production data from a cast iron foundry, despite the limitation of complete casting traceability. The aim is to help sand foundries to produce good castings. A reference foundry, which produces mainly automotive and oven parts with automatic sand molding and pouring machines, was selected. Past data, where only good castings were produced, were extracted from the database to determine parameter control limits (upper and lower control limits) with the aid of statistical approach. To identify critical process parameters associated with casting defects, process data from the zero and high scrap production batches were systematically compared. This method clearly identified unstable parameters without exact synchronization between inline and part quality data. Molding sand moisture, temperature and compactability, liquidus temperature of the melt, phosphorus content, carbon equivalent and pouring temperature were found to be the critical parameters to be stabilized. Finally, a regression model for predicting and controlling of molding sand moisture and liquidus temperature of the melt was created. The determined boundaries and the models were helpful for the foundry in assisting ongoing production control and correction of process inputs to achieve target casting quality.

Utilization of Kotter's Stages and Statistical Process Control to Implement and Sustain Delirium Screening in PICU.

We utilized Kotter's stages of transformation as guiding blocks for implementing change and undertook simultaneous efforts to decrease delirium rates. In addition, we used statistical process control to monitor screening compliance and delirium rates. Education on delirium assessment and prevention created a sense of urgency among nursing and medical staff (Stage #1). Stakeholder analysis and a key driver diagram were presented to leadership to create a guiding coalition (Stage #2). Process mapping and failure mode and effect analysis created a vision and strategy for the project (Stage #3). Multiple methods were utilized to communicate the change vision (Stage #4) and empower broad-based action (Stage #5). We celebrated small incremental achievements or short-term wins (Stage #6) by reporting delirium compliance and rate on a control chart. We completed 5837 delirium screens on 763 patients (6689 opportunities). The median delirium screening compliance rate was 87.2% (13.4% positive screens). Small achievements produced by the project were communicated to staff, propelling future efforts to recognize and prevent delirium. This momentum was a driver for more change (Stage #7) to ensure a long-term cultural change in the unit (Stage #8). Kotter's principle of change management, along with an emphasis on data-driven process control, may result in sustained improvement.

Plant-wide interoperability and decoupled, data-driven process control with message bus communication

Conventional industrial communication systems suffer from rigidness, inflexibility and lack of scalability. The environment is heterogeneous as the systems exchange data with a variety communication protocols, some of which are proprietary. This makes it laborious and expensive to reconfigure or upgrade the systems. As the solution, this article proposes a message-bus-based communication architecture to enable information exchange between systems regardless of their geographical location and position within the functional hierarchy of the plant. The architecture not only enables communication to cross the conventional physical borders but also provides scalability to growing data volumes and network sizes. As proofs of concept, the article presents a prototype in three environments: a copper smelter, a steel plant and a distillation column. The results suggest that the message-bus-based approach has potential to renew industrial communications, a core part of the fourth industrial revolution.

Data-driven process control for manufacturing spiral bevel and hypoid gears by using design for six sigma (DFSS) considering numerical loaded tooth contact analysis (NLTCA)

Data-driven process control considering both geometric and loaded contact performance evaluations has been an increasingly important stage in field of spiral bevel and hypoid gears. A new data-driven manufacturing process control strategy is proposed for a high performance spiral bevel and hypoid gears. Here, to distinguish with the conventional simulated loaded tooth contact analysis (SLTCA) using economical finite element software package, the numerical loaded tooth contact analysis (NLTCA) is of more flexibility and practicality. In light of the advantages of the improved design for six sigma (DFSS), it is integrated with NLTCA for establishing a novel data-driven process control of gear manufacturing. Firstly, in improved DFSS framework, quality function deployment (QFD) is used to determine four sub-objective high-performance evaluation items. Then, their data-driven relationships between machine settings are respectively determined by using NLTCA. In particular, the manufacturing process control is further converted into multi-objective optimization (MOO) modification of the hypoid generator settings. Finally, an interactive preference point approach is applied for data-driven control of its iterative step and it can obtain a robust solution from Pareto optimal front. A case study is provided to verify the proposed methodology.

A survey on data-driven process monitoring and diagnostic methods for variation reduction in multi-station assembly systems

PurposeDriven by the development in sensing techniques and information and communications technology, and their applications in the manufacturing system, data-driven quality control methods play an essential role in the quality improvement of assembly products. This paper aims to review the development of data-driven modeling methods for process monitoring and fault diagnosis in multi-station assembly systems. Furthermore, the authors discuss the applications of the methods proposed and present suggestions for future studies in data mining for quality control in product assembly.Design/methodology/approachThis paper provides an outline of data-driven process monitoring and fault diagnosis methods for reduction in variation. The development of statistical process monitoring techniques and diagnosis methods, such as pattern matching, estimation-based analysis and artificial intelligence-based diagnostics, is introduced.FindingsA classification structure for data-driven process control techniques and the limitations of their applications in multi-station assembly processes are discussed. From the perspective of the engineering requirements of real, dynamic, nonlinear and uncertain assembly systems, future trends in sensing system location, data mining and data fusion techniques for variation reduction are suggested.Originality/valueThis paper reveals the development of process monitoring and fault diagnosis techniques, and their applications in variation reduction in multi-station assembly.

Dynamics-inspired feature extraction in semiconductor manufacturing processes

Abstract The ability to exploit data-driven process control and decision making frameworks is rapidly becoming critical to success in semiconductor manufacturing. At the same time, advances in manufacturing equipment sensors has seen dramatic increases in sampling rates in recent years, which has led to the ability to capture transients effects in signals with higher fidelity than previously possible. It is known that data-driven process control and decision making methodologies rely on the process of extraction of useful information from raw data signals. To that end, the current manuscript presents a novel methodology for extraction of information from data in the form of a feature set that faithfully and reliably depicts both the transient and stationary portions of the signals. The solution proposed is an automated dynamics-inspired approach that looks to segment a signal into steady state and transient components before summarizing each segment into a set of relevant signatures. The steady state segments are summarized through a set of statistics and each transient is reduced to a set of parameters relating to the underlying system dynamics, such as settling time, rise time, overshoots, etc. The impactful novel information content of the resulting dynamics-inspired feature set is evaluated by application to chamber matching, product defect level prediction and product quality characteristic prediction in etch and deposition processes executed in various tools across several modern 300 mm fabs.

Quantisation and data quality: Implications for system identification

In the pursuit of online, data-driven process control, there is a need to determine the quality of the data being processed before actually using it. One area that needs to be considered is data quantisation. Although in many applications it has been assumed that the impact of quantisation is to solely increase the variance of the signal, in certain cases this may not hold. This is especially the case when dealing with signals from poorly quantised sources, such as temperature sensors. In this case, the effect of quantisation cannot be solely considered by the impact of the increase in the variance. Therefore, this paper will examine the effects of small scale quantisation with the view of determining an appropriate metric for measuring the effect on data quality of quantisation. It will be shown that if the ratio of the unquantised signal variance and the distance between quantisation step sizes are below a given threshold, then the identification of the process parameters will be problematic. Detailed numerical simulations as well as an example drawn from a real system are presented to validate the proposed metrics and approach.