Figure - available from: The International Journal of Advanced Manufacturing Technology
This content is subject to copyright. Terms and conditions apply.
![Grinding burn limit of preliminary experiments adapted from [28]](publication/341433890/figure/fig7/AS:892045770100739@1589691693130/Grinding-burn-limit-of-preliminary-experiments-adapted-from-28.png)
Source publication
In this study, self-optimization of a grinding machine is demonstrated with respect to production costs, while fulfilling quality and safety constraints. The quality requirements of the final workpiece are defined with respect to grinding burn and surface roughness, and the safety constrains are defined with respect to the temperature at the grindi...
Similar publications
In this study, self-optimization of a grinding machine is demonstrated with respect to production costs, while fulfilling quality and safety constraints. The quality requirements of the final workpiece are defined with respect to grinding burn and surface roughness, and the safety constraints are defined with respect to the temperature at the grind...
Citations
... Since self-optimization of a process is not supposed to have a negative impact on the quality of this process, standard learning methods are not applicable. In contrast to just focusing on minimizing the required number of realworld experiments (Maier et al., 2020), an optimization with constraints (Rattunde et al., 2021) is more appropriate for industrial standards. Such safe learning methods become particularly relevant for real-world interacting applications of machines (Chatzilygeroudis et al., 2018). ...
Intelligent manufacturing applications and agent-based implementations are scientifically investigated due to the enormous potential of industrial process optimization. The most widespread data-driven approach is the use of experimental history under test conditions for training, followed by execution of the trained model. Since factors, such as tool wear, affect the process, the experimental history has to be compiled extensively. In addition, individual machine noise implies that the models are not easily transferable to other (theoretically identical) machines. In contrast, a continual learning system should have the capacity to adapt (slightly) to a changing environment, e.g., another machine under different working conditions. Since this adaptation can potentially have a negative impact on process quality, especially in industry, safe optimization methods are required. In this article, we present a significant step towards self-optimizing machines in industry, by introducing a novel method for efficient safe contextual optimization and continuously trading-off between exploration and exploitation. Furthermore, an appropriate data discard strategy and local approximation techniques enable continual optimization. The approach is implemented as generic software module for an industrial edge control device. We apply this module to a steel straightening machine as an example, enabling it to adapt safely to changing environments.
... As a result, less experimental data are required to obtain a model of the manufacturing process. Maier et al. [62] successfully applied this method to optimize grinding parameters after a few trials for an unknown workpiece and tool combination. Table 2 summarizes the mentioned process modelling methods. ...
Industrial control systems play a central role in today’s manufacturing systems. Ongoing trends towards more flexibility and sustainability, while maintaining and improving production capacities and productivity, increase the complexity of production systems drastically. To cope with these challenges, advanced control algorithms and further developments are required. In recent years, developments in Artificial Intelligence (AI)-based methods have gained significantly attention and relevance in research and the industry for future industrial control systems. AI-based approaches are increasingly explored at various industrial control systems levels ranging from single automation devices to the real-time control of complex machines, production processes and overall factories supervision and optimization. Thereby, AI solutions are exploited with reference to different industrial control applications from sensor fusion methods to novel model predictive control techniques, from self-optimizing machines to collaborative robots, from factory adaptive automation systems to production supervisory control systems. The aim of the present perspective paper is to provide an overview of novel applications of AI methods to industrial control systems on different levels, so as to improve the production systems’ self-learning capacities, their overall performance, the related process and product quality, the optimal use of resources and the industrial systems safety, and resilience to varying boundary conditions and production requests. Finally, major open challenges and future perspectives are addressed.
... Numerous acquisition functions are presented in the BO literature (Hernández-Lobato et al., 2016;Garrido-Merchán and Hernández-Lobato, 2019;Gardner et al., 2014). BO has been successfully used in numerous applications, such as manufacturing (Maier et al., 2020;Guidetti et al., 2021) or control under safety constraints (Khosravi et al., 2022). In this work, we use the BO algorithm studied in (Guidetti et al., 2022), that was specifically designed for the configuration of advanced manufacturing processes such as FFF. ...
The tuning of fused filament fabrication parameters is notoriously challenging. We propose an autonomous data-driven method to select parameters based on in situ measurements. We use a laser sensor to evaluate the surface roughness of a printed part. We then correlate the roughness to the mechanical properties of the part, and show how print quality affects mechanical performance. Finally, we use Bayesian optimization to search for optimal print parameters. We demonstrate our method by printing liquid crystal polymer samples, and successfully find parameters that produce high-performance prints and maximize the manufacturing process efficiency.
... All these approaches require a large number of samples, to either achieve good predictive capabilities, or to reliably cover all possible process variations. Previous work on data-driven optimization of manufacturing processes [10], [11] has shown that Bayesian optimization (BO) is an information efficient and effective technique to automate the configuration of industrial processes with a limited number of experiments. BO has been applied in parallel or nested formulations for the tuning of algorithms [12], the optimization of black-box functions [13], [14], [15], or the tuning of cascaded controllers [16]. ...
We propose a framework for the configuration and operation of expensive-to-evaluate advanced manufacturing methods, based on Bayesian optimization. The framework unifies a tailored acquisition function, a parallel acquisition procedure, and the integration of process information providing context to the optimization procedure. The novel acquisition function is demonstrated, analyzed and compared on state-of-the-art benchmarking problems. We apply the optimization approach to atmospheric plasma spraying and fused deposition modeling. Our results demonstrate that the proposed framework can efficiently find input parameters that produce the desired outcome and minimize the process cost.
... Klippel et al. [18] trained a multi-layer perceptron to predict the cutting force of the orthogonal machining process using training data generated from a smoothed particle hydrodynamics simulation. Maier et al. [19] investigated Bayesian-based ML techniques for the probabilistic prediction and optimization of process indicators such as cutting forces and roughness. However, the application of data-driven methods in industry is largely restricted by the available datasets as well as the robustness and generalizability of the training models. ...
Condition monitoring of manufacturing processes plays a vital role in the era of Industry 4.0 and sustainable development goals. The use of condition monitoring enables soft sensing of indirectly measurable or non-measurable production indicators, e.g., cutting forces, as the foundation for a digital shadow or digital twin powered by emerging technologies such as artificial intelligence (AI) and edge computing techniques. Hence, when properly applied, it leads to an improvement in process understanding, production planning, quality control, and predictive maintenance. Despite these merits, the adaptability and reliability of condition monitoring at the production field level are limited by the modeling uncertainties due to the rapidly changing industrial environment driven by dynamic customer demands. This study presents a novel strategy to overcome these shortcomings by employing machine internal signals and prior knowledge-embedded machine learning techniques. A concept of incorporating production domain-specific physical models and logical rules into generalized data-driven methods is proposed. A hybrid learning force model based on machine internal signals was constructed. As a result, the soft sensor encompasses stochastic effects that are neglected in the existing model and estimates the reliability of the modeled findings. Prospectively, this allows for online parametrization of the model within dynamically changing conditions. This concept is exemplarily implemented in an industrial use case.
... This project has been funded by the Swiss Innovation Agency (grant №37896) and by the Swiss National Science Foundation under NCCR Automation (grant №180545). 1 Previous work on data-driven optimization of manufacturing processes [8], [9] has shown that Bayesian optimization (BO) is an information efficient and effective technique to automate the configuration of industrial processes with a limited number of experiments. BO has been applied in parallel or nested formulations for the tuning of algorithms [10], the optimization of black-box functions [11]- [13], or the tuning of cascaded controllers [14]. ...
... calculate I(x) of all x ∈ U , make a virtual copy Tv ← T ;8 for i = 1 to n do9 Using the the data in Tv, model the constraints c(·); 10 Use (6) to calculate FP(x) of all x ∈ U ;11 Select the candidate x * i using Alg. 1;12 Remove x * i from U and add it to B;13 Expand Tv with (x * i , µc(x * i )), where µc(x * i ) collects the predictive means of the constraints c(α FIP (x * i ) < for at least half of the candidates x * i selected for a batch B ; 17 return feasible stress index minimizer x + = arg min x∈T f S(x), ...
We propose a framework for the configuration and operation of expensive-to-evaluate advanced manufacturing methods, based on Bayesian optimization. The framework unifies a tailored acquisition function, a parallel acquisition procedure, and the integration of process information providing context to the optimization procedure. The novel acquisition function is demonstrated and analyzed on benchmark illustrative problems. We apply the optimization approach to atmospheric plasma spraying in simulation and experiments. Our results demonstrate that the proposed framework can efficiently find input parameters that produce the desired outcome and minimize the process cost.
... Optimization constraints can be modeled as separate GPs [24], and the inputs corresponding to fulfilling the constraints with high probability are chosen by adapting the acquisition function. The approach has been demonstrated for manufacturing processes [25], [26], and for the run-to-run control of reluctance actuators [27]. Another approach to respect safety constraints, known as safe Bayesian optimization (SafeOpt) provides probabilistic guarantees that all candidate samples remain in the constraint set [28]. ...
This article presents an automated, model-free, data-driven method for the safe tuning of PID cascade controller gains based on Bayesian optimization. The optimization objective is composed of data-driven performance metrics and modeled using Gaussian processes. The safety requirement is imposed via a barrier-like term in the objective, which is introduced to account for operational changes in the system. We further introduce a data-driven constraint that captures the stability requirements from system data. Numerical evaluation shows that the proposed approach outperforms relay feedback autotuning and quickly converges to the global optimum, thanks to a tailored stopping criterion. We demonstrate the performance of the method through simulations and experiments. For experimental implementation, in addition to the introduced safety constraint, we integrate a method for automatic detection of the critical gains and extend the optimization objective with a penalty depending on the proximity of the current candidate points to the critical gains. The resulting automated tuning method optimizes system performance while ensuring stability and standardization.
... To improve the current optimization approaches, [35] demonstrated self-optimization of a grinding process by using adaptive sampling combined with Gaussian process models. Gaussian process models are flexible and provide an uncertainty estimation of the prediction, which is not provided by neural networks. ...
... In this way, optimal parameters are determined with a few experiments. Fig. 4. Gaussian process models combined with adaptive sampling from [35]. Left side presents in green the 95% confidence interval and the acquisition function, the maximum of which gives the new parameters for the next experiment. ...
... The discussed process optimization methods neglect the information made available during the set-up process, counting either on prior knowledge within the learned models, or on statistical insights from the collected data. Previous work on data-driven optimization of manufacturing processes [9], [10] has shown that Bayesian optimization (BO) is an information efficient and effective technique to automate the set-up of industrial processes in a limited amount of experiments. ...
Recent work has shown constrained Bayesian optimization to be a powerful technique for the optimization of industrial processes. We adapt this framework to the set-up and optimization of atmospheric plasma spraying processes. We propose and validate a Gaussian process modeling structure to predict coatings properties. We introduce a parallel acquisition procedure tailored on the process characteristics and propose an algorithm that adapts to real-time process measurements to improve reproducibility. We validate our optimization method numerically and experimentally, and demonstrate that it can efficiently find input parameters that produce the desired coating and minimize the process cost.
... Maier, M. et al. [37] implemented data-driven optimization methods for studying the grinding process. In particular, a self-optimization algorithm of a grinding machine was developed for decreasing production costs. ...
The intensifying of the manufacturing process and increasing the efficiency of production planning of precise and non-rigid parts, mainly crankshafts, are the first-priority task in modern manufacturing. The use of various methods for controlling the cutting force under cylindrical infeed grinding and studying its impact on crankpin machining quality and accuracy can improve machining efficiency. The paper deals with developing a comprehensive scientific and methodological approach for determining the experimental dependence parameters' quantitative values for cutting-force calculation in cylindrical infeed grinding. The main stages of creating a method for conducting a virtual experiment to determine the cutting force depending on the array of defining parameters obtained from experimental studies are outlined. It will make it possible to get recommendations for the formation of a valid route for crankpin machining. The research's scientific novelty lies in the developed scientific and methodological approach for determining the cutting force, based on the integrated application of an artificial neural network (ANN) and multi-parametric quasi-linear regression analysis. In particular, on production conditions, the proposed method allows the rapid and accurate assessment of the technological parameters' influence on the power characteristics for the cutting process. A numerical experiment was conducted to study the cutting force and evaluate its value's primary indicators based on the proposed method. The study's practical value lies in studying how to improve the grinding performance of the main bearing and connecting rod journals by intensifying cutting modes and optimizing the structure of machining cycles.
