BUSCADOR RECOLECTA

The recent convergence of a group of technological advances including the Internet of Things, data analytics, robotics and advanced simulation environments, in addition to the exponential growth experienced by Artificial Intelligence in the last few years, has led to the emergence of a new paradigm in manufacturing: Industry 4.0. This paradigm is characterized by the digitalization of production processes and revolves around the concept of cyber-physical systems, in which real-world processes are integrated with those that take place in the virtual space. This type of system is enabling the creation of more efficient, flexible and adaptive industrial environments.

However, the rapid pace at which all these transformations are occurring has made it difficult for some stakeholders to assimilate the new manufacturing paradigm due to the lack of a defined roadmap for its adoption. Although certain initiatives have defined reference architectures, most of them present the limitation of focusing only on theoretical aspects without delving into practical issues. The main objective of this thesis is the design and development of methods for automatic reconfiguration and supervision of cyber-physical systems by means of digital twins and Artificial Intelligence. The scientific-technical challenge lies in the fact that these methods must be materialized in Artificial Intelligence modules, embedded in computational platforms, ready to be applied to production processes in order to ease the digitalization of manufacturing companies.

The first module is based on an automated machine learning methodology that allows to generate optimal parametrizations of production processes through a data-driven approach. In this methodology, starting from a set of historical data that must include variables of the production process, as well as the Key Performance Indicators to be optimized, a series of stages are carried out to pre-process the data; select the most relevant variables for each Key Performance Indicator; determine between a series of algorithms with optimized hyperparameters and pre-established modelling strategies which is the most suitable one for modelling each Key Performance Indicator; and finally, use the selected models to solve a multi-objective optimization problem, obtaining as a result the optimal parametrizations of the production process. The capabilities of this module make it suitable to operate as the intelligence layer of a digital twin. This was demonstrated by integrating it into a digital twin of a pilot line. As results, models with values for the coefficient of determination higher than 0.98 were obtained and optimal Pareto sets were generated according to different constraints. Additionally, the digital twin proved to be able to co-evolve with the production process when it was deliberately exposed to different types of disturbances that changed its behavior.

The second module has the objective of maintaining the productivity of manufacturing equipment using a Fuzzy Logic strategy, specifically a fuzzy inference system. This module can be applied to heterogeneous equipment and has great relevance since it allows in a simple way to avoid fluctuations in its working speed, preventing the under-utilization or over-utilization of the equipment once it has been configured for a desired regime. The module is capable of operate within a certain range of fluctuations, whereas, if this range is exceeded, it is assumed that an abnormal situation has occurred and, therefore, the production should be stopped to avoid undesirable consequences. This module was also tested in a pilot line, demonstrating its effectiveness.

The development of this thesis has sought to contribute to the democratization of Artificial Intelligence in the industrial sector by providing different stakeholders with tools that allow building Artificial Intelligence-based solutions in a simple and affordable way.

RESUMEN

La reciente convergencia de un grupo de avances tecnológicos como el Internet de las Cosas, la analítica de datos, la robótica y los entornos avanzados de simulación, unido al rápido crecimiento de la Inteligencia Artificial, han propiciado el surgimiento de un nuevo paradigma de fabricación: la Industria 4.0. Este se caracteriza por la digitalización de los procesos productivos y gira en torno al concepto de sistema ciberfísico, en el cual se integran los procesos del mundo real con los del espacio virtual. Estos sistemas permiten la creación de entornos industriales más eficientes, flexibles y adaptables.

Sin embargo, el ritmo vertiginoso al que están ocurriendo estas transformaciones dificulta la asimilación del nuevo paradigma de fabricación a ciertos actores ya que no existe una ruta prestablecida para su adopción. Si bien algunas iniciativas han definido arquitecturas de referencia, estas presentan la limitación de enfocarse solamente en aspectos teóricos sin profundizar en cuestiones prácticas. Esta tesis tiene como objetivo principal el diseño y desarrollo de métodos de reconfiguración y supervisión automática de sistemas ciberfísicos mediante gemelos digitales e Inteligencia Artificial. El reto científico-técnico radica en que estos métodos deben materializarse en módulos de Inteligencia Artificial, embebidos en plataformas computacionales, listos para ser aplicados a procesos productivos con el fin de facilitar la digitalización en las empresas de fabricación.

El primer módulo se basa en una metodología de aprendizaje automático automatizado para generar parametrizaciones óptimas de los procesos productivos. A partir de datos históricos que deben incluir variables del proceso productivo y los Indicadores Clave de Desempeño que se desean optimizar, se llevan a cabo una serie de etapas para preprocesar los datos; seleccionar las variables con mayor influencia para cada Indicador Clave de Desempeño; determinar de entre una serie de algoritmos y estrategias de modelado cuál es el más adecuado para modelar cada Indicador Clave de Desempeño y; finalmente, utilizar los modelos seleccionados para plantear un problema de optimización multiobjetivo cuya solución son las parametrizaciones óptimas del proceso productivo. Las principales ventajas de este método son que, a pesar de su complejidad, su utilización es sencilla y puede ser aplicado a un amplio rango de procesos productivos. Las capacidades de este módulo le permiten actuar como capa de inteligencia de gemelos digitales, lo cual fue demostrado en una línea piloto. Como resultados se obtuvieron modelos con valores del coeficiente de determinación superiores a 0.98 y se generaron conjuntos de Pareto óptimos para diferentes restricciones. Además, el gemelo digital demostró ser capaz de coevolucionar con el proceso productivo al ser sometido deliberadamente a varias perturbaciones que alteraban su comportamiento.

El segundo módulo tiene el objetivo de mantener la productividad de los equipos de fabricación utilizando una estrategia basada en la Lógica Borrosa, específicamente en los sistemas de inferencia borrosa. Este módulo se puede aplicar a equipos de carácter heterogéneo y tiene gran relevancia ya que permite de una manera sencilla evitar fluctuaciones en su ritmo de trabajo, previniendo la subutilización o sobreutilización de los equipos una vez que han sido configurados para un régimen de trabajo. El módulo es capaz de operar dentro de un determinado rango de fluctuaciones, mientras que, si se sobrepasa este rango, se asume que ha ocurrido una situación anómala y se detiene la producción para evitar consecuencias no deseadas. La efectividad de este módulo también fue probada en una línea piloto.

Con el desarrollo de esta tesis se busca contribuir a la democratización de la Inteligencia Artificial en el sector industrial al facilitar a los diferentes actores herramientas para construir soluciones basadas en Inteligencia Artificial de manera simple y asequible.

Optimization on the basis of sustainability brings important benefits to manufacturing process as sustainable productions constitute a crucial aspect in modern manufacturing. This paper presents a new formalized framework for optimizing the sustainability of manufacturing processes. Unlike previous approaches, the proposed technique combines a methodology for selecting the sustainability indicators and a multi-objective optimization for improving the three sustainability pillars (economy, environment and society). While selecting the significant sustainability indicators in the considered manufacturing process relies on the ABC judgment method, the Saaty's method enables weighting the chosen indicators in order to combine them into suitable economic, environmental and social sustainability indexes. Other technological aspects, usually taken as objectives in previous works, are considered constraints in the proposed approach. The optimization is performed by using nature inspired heuristics, which return the set of non-dominated solutions (also known as Pareto front), from which the most convenient alternative is chosen by the decision maker, depending on the specific conditions of the process. For illustrating the usage of the proposed framework, it is applied to the optimization of a submerged arc welding process. Compared with currently used welding parameters, the computed optimal solution outperforms the economic and environmental sustainability while keeps equal the social impact. The results show not only the effectiveness of the proposed approach, but also its flexibility by giving a set of possible solutions which can be chosen depending on how are ranked the sustainability pillars., This work was supported in part by the project Power2Power: Providing Next-Generation Silicon-Based Power Solutions in Transport and Machinery for Significant Decarbonisation in the Next Decade, funded by the Electronic Component Systems for European Leadership (ECSEL-JU) Joint Undertaking and MICINN under Grant 826417, in part by the European Commission through the Project H2020 Platform enable KITs of Artificial Intelligence for an Easy Uptake of SMEs (KITT4SME) under Grant 952119, and in part by the Cuban National Program on Basic Sciences through the Project Multi-Objectives Optimization Heuristics for Industrial Applications, under Grant P223LH001-068., Peer reviewed

Quality is defined as the extent to which a product conforms to the design specifications and how it complies with the requirements of component functionality. For some industries, such as automotive and aeronautical, the quality of their parts is very important given the high requirements to which they are subject. However, difficulties arise from the fact that a measure of quality can only be evaluated ‘‘out-of-process”, resulting in losses because there is no alternative to removing defective parts from the production line. Therefore, it is necessary to apply Artificial Intelligence-based kits/solutions that provide in-process estimation to predict quality from some measured variables.

The main goal of these datasets is to monitor the final quality of the manufactured components or parts by estimating surface roughness from vibration signals and cutting parameters information using Artificial Intelligence-based solutions. Surface roughness is an essential feature in quality control defined by the deviation in the direction of the normal vector of a real surface from its ideal form. Because the roughness measurement is an offline and post process procedure, being able to estimate this value online brings a series of benefits in terms of time and cost reduction in manufacturing lines, energy efficiency, unnecessary wear of tools and machines, etc. Once a part has been detected with a surface quality below what is desired, a series of corrective measures can be applied for the following operations, such as: reducing the feed rate percentage, increasing the percentage of spindle speed or reducing the axial depth per pass, etc., European Commission: KITT4SME - platform-enabled KITs of arTificial intelligence FOR an easy uptake by SMEs (952119)., Peer reviewed

In this modern era where Industry 4.0, plays a crucial role in enhancing productivity, quality, and resource utilization by digitalizing and providing smart operation to industrial systems. Therefore, there is a need to establish a framework that enhances productivity and quality of work to achieve the net-zero from industry. In this study, a comprehensive and generic analytical framework has been established to mitigate or lessen the research and technological gap in the manufacturing sector. In addition to that, the key stages involved in artificial intelligence (AI) based modelling and optimization analysis for manufacturing systems have also been incorporated. To assess the proposed AI framework, electric discharge machining (EDM) as a case study has been selected. The focus enlightens the emergence of optimizing the material removal rate (MRR) and surface roughness (SR) for Inconel 617 (IN617) material. A full factorial design of the experiment was carried out for experimentation. After that, an artificial neural network (ANN) as a modelling framework is selected, and fine-tuning of hyperparameters during training has been carried out. To validate the predictive performance of the trained models, an external validation (Valext) test has been conducted. Through sensitivity analysis (SA) on the developed AI framework, the most influential factors affecting MRR and SR in EDM have been identified. Specifically, powder concentration (Cp) contributes the most to the percentage significance, accounting for 79.00 % towards MRR, followed by treatment (16.35 %) and 4.67 % surfactant concentration (Sc). However, the highest % significance in SR is given by Sc (36.86 %), followed by Cp (33.23 %), and then treatment (29.90 %), respectively. Furthermore, a parametric optimization has been performed using the framework and found that MRR and SR are 93.75 % and 58.90 % better than experimental data. This successful performance optimization proposed by the framework has the potential for application to other manufacturing systems., The authors appreciate the support from Researchers Supporting Project number (RSPD2023R702), King Saud University, Riyadh, Saudi Arabia. Authors wish also to thank projects H2020 KITT4SME GA 952119, and MICINN and NextGenerationEU/PRTR SELFRECO grant ID PID2021-127763OB-100., Peer reviewed