http://bura.brunel.ac.uk/handle/2438/29517 Theses for the Special Research Institutes, BCAST 2026-04-27T23:08:11Z http://bura.brunel.ac.uk/handle/2438/33202 Title: Modelling of precipitation hardening during non-isothermal thermomechanical treatment of 6 series aluminium alloys Authors: Gharavian, Somayeh Abstract: The future of the automotive industry can be viewed as contingent upon the further development of aluminium alloys. This can be primarily achieved when the behaviour of aluminium alloys during the thermomechanical treatment process of hardening is comprehensively understood and predicted. This study focuses on developing a comprehensive mathematical model for predicting the mechanical behaviour of Al-Mg-Si(Cu) systems subjected to non-isothermal heat treatment to ultimately enable the prediction of mechanical behaviour in the form of a software tool. The Kampmann and Wagner numerical model is among the well-studied mathematical models for precipitation hardening; this model was adapted as the base model for this study where it was further incorporated with critical factors such as multi-stage aging, clustering effects, and the influence of plastic deformation. By coupling the framework to a thermomechanical database and refining precipitation kinetics, the model exhibited improved accuracy in simulating the evolution of microstructure and the mechanical properties under industry specific conditions. Validation of the developed model was carried out by comparing with experimental data obtained from laboratory experiments on Al-Mg-Si (Cu) alloys. These experiments included varying heat treatment duration, temperatures, plastic deformation and different cooling/heating rate to replicate the industrial conditions. The results of the model demonstrate the capability to predict multi-stage aging processes under non-isothermal conditions, facilitating the analysis of various quenching and heating rates. A key advantage is its integration of precipitation and clustering predictions within a unified framework, enabling accurate assessments across a broad range of aging temperatures from natural aging to elevated temperatures like 200°C. Furthermore, the model incorporates the effects of plastic deformation in the form of 4–8% cold stretching, enabling the exploration of not only work hardening but also the influence of deformation on the thermodynamics and kinetics of the process. These findings highlight the significant potential of mathematical modelling to optimize heat treatment process design, substantially reducing the workload in the automotive industry. Moreover, the model has shown significant potential to be used as a helpful tool towards alloy design purposes with further development and validation with experimental data. Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London 2024-01-01T00:00:00Z http://bura.brunel.ac.uk/handle/2438/33188 Title: Rationalisation of aluminium alloys using machine learning and Artificial Intelligence Authors: Tiwari, Tanu Abstract: Aluminium alloys are widely used across various sectors of engineering due to their lower density combined with higher strength compared to many existing alloys of other metals. These unique characteristics have led to an increased demand for and discovery of new aluminium alloys with targeted properties and compositions. Traditional methods of designing new mate-rials with desired properties, such as trial-and-error and reliance on domain experts' experience, are time-consuming and expensive. These techniques also expand the search area for suitable alloys. In this research, we propose a machine learning-based design system to reduce the number of grades across all series of age-hardenable and non-age-hardenable aluminium alloys. The sys-tem collects features based on chemical composition, mechanical properties, corrosion re-sistance, weldability, and thermal and electrical properties under different tempering and hard-ening conditions for machine learning modelling. A combination of PCA (Principal Compo-nent Analysis) and K-means clustering is applied for clustering and sub-clustering similar al-loys based on their compositional and property profiles into clusters and sub-clusters. Next, an optimisation algorithm, namely a multi-property decision-making method, i.e., TOPSIS (Tech-nique for Order Preference by Similarity to Ideal Solution), identifies the optimum alloys within each sub-cluster. These selected alloys exhibit a balanced set of properties that effec-tively represent the range of characteristics found among other alloys in the same sub-cluster. Subsequently, a recycling algorithm is applied to predict the mixing ratio based on closeness scores generated by the optimisation algorithm. This process mixes the optimum alloy in each sub-cluster with the remaining alloys in the sub-cluster, resulting in a single optimised alloy as determined by the optimisation algorithm. This method significantly reduces the number of alloy grades while maintaining key material properties and enhancing recyclability, which has a metallurgical basis. This design system is enhanced and developed into a dedicated recycling software application, offering a practical tool for the aluminium industry. It supports sustainable development and improves recycling efficiency, aligning alloy manufacturing with the principles of the circular economy. Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London 2025-01-01T00:00:00Z http://bura.brunel.ac.uk/handle/2438/33079 Title: Rationalisation of steel grades and specifications using machine learning techniques Authors: Sadegh, Jalalian Abstract: There are an excessive number of steel grades currently in use. However, many of them are used in the same application despite differences in chemical composition and processing conditions, and in some cases shows equivalent ranges of properties. These huge number of grades poses challenges for sustainable recycling and increases production complexity and cost. This study introduces a multi-phase, application-driven framework to simplify the steel grade system and reduce the number of grades by proposing a novel approach called K-Means Reduction Process (KMRP). The framework was applied to 148 carbon and 288 stainless steel grades, including chemical composition, processing conditions, and mechanical properties (hardness, UTS, YS, and elongation). Machine learning models were first used to quantify the influence of alloying elements and processing conditions on mechanical performance. K-Means clustering was then applied to group grades based on performance to identify steels that shared equivalent property profiles, with four distinct clusters identified including ferritic/low-carbon steels, medium-carbon and martensitic steels, high-carbon steels, and austenitic steels. These clusters revealed significant redundancy, with multiple grades from existing steel classifications occupying the same mechanical property space. In the reduction phase, KMRP identified the minimal set of grades required to preserve full mechanical property coverage within the generated clusters. Two sustainability-driven strategies were implemented: (1) tramp-element avoidance, favouring grades with low Cu and Sn, and (2) tramp-element tolerance, prioritising grades compatible with scrap-based recycling. While both approaches reduced reliance on critical raw materials (Mo, Ni, V, Ti), this study focused on the tramp-tolerance strategy as the most relevant for advancing circular economy objectives. Under this approach, the number of carbon steel grades were reduced by 38.4% (from 146 to 90) and stainless steel grades by 52.8% (from 288 to 136), while fully preserving the original mechanical property ranges, including UTS ranges of 295–2450 MPa for carbon steels and 120–1970 MPa for stainless steels, and elongation ranges of 7–41% and 2–55%, respectively. These results demonstrated that KMRP can successfully simplify the steel grade system while supporting circularity, reducing dependency on critical elements, and improving the sustainability of future steel production. Moreover, the methodology is generalisable and can be applied to other domains where reducing redundant options is essential, such as pharmaceutical applications. Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London 2025-01-01T00:00:00Z http://bura.brunel.ac.uk/handle/2438/32328 Title: Machine learning for alloy design: Interpolation, extrapolation, and dataset integration in the prediction of 6xxx-series aluminium alloy properties Authors: Birchall, Adam Abstract: The optimisation of alloy compositions and processing has always been a goal in aluminium alloy design, and the recent increase in volume of data generated has enabled the use of data driven approaches to attempt to aid in this area. The potential for accurate predictions of mechanical properties without the need for lengthy and costly experiments hails as a means to quickly evaluate large areas of the design space. Machine learning models coupled with optimisation methods pose a method of finding sets of input parameters to achieve tailored properties. These methods however are highly data-dependent and in the domain of alloy design, obtaining data detailing alloy compositions, processing conditions and properties can be a costly affair. This work introduces 3 novel datasets to attempt to complete such work on and contributes to the availability of data in the domain for further research. An opensourced based dataset and synthetic dataset were curated and curated, and can be used by all for further research, a proprietary dataset was also curated with the support of an industry partner but is not publicly available. These three datasets all include compositional and thermomechanical processing variance and property values of corresponding alloy yield tensile strength values. Yield strength was used as a target property for the majority of this work due to its relative simplicity as a tensile property, less impacted by plastic deformation behaviour than other properties, but also its value as a property used to define material use. These datasets include an open-source dataset consisting of low quality, highly varied data, a proprietary dataset consisting of accurate measurements of a small part of the domain, and a synthetic dataset generated from a physics-based model, used in later experiments. Interpolation has been shown to be a relatively straightforward task in literature time and time again, with accurate predictions from decision tree based models dominating this type of problem. This work demonstrates their use can effectively predict yield strength values within the domain of existing data to an error of 19.5MPa on proprietary datasets detailing 6000 series aluminium alloys, with typical yield strength values of 350MPa. Similar models are able to achieve close results. These models however lack insight into the aluminium domain, and are frequently subject to the biases of the data which trained them. Qualitative results go further and show that the responses of these models can be unphysical. This work shows that the use of physical metallurgy variables through averaging methods, commonly used to attempt to aid with domain knowledge inclusion in literature, demonstrates no improvement to model accuracies, contributing to the discussion of this method’s use in similar work. Investigations into the extrapolation capabilities showed that typical models used in the domain demonstrate poor extrapolation performances, a metric not commonly highlighted currently in the domain despite the common purpose of models to guide experiment to outperforming alloys. The best performing typical model for extrapolation being linear regression, achieving errors of 57MPa on 6000 series aluminium alloys. The introduction of Delta modelling techniques taken from other machine learning domains and applied to this property prediction was able to reduce these extrapolation errors to 51MPa however a significant margin, the best performing method for improving extrapolation capability was a transfer learning derivative, domain adaptation. Applying domain adaptation techniques to the datasets compiled and curated in this work coupled with a novel implementation of delta modelling via the introduced ’Composite model’ further reduced these extrapolation errors by 31% from previous best models. Multiple datasets, with varying quality and differing input feature sets, were used in a composite model leveraging domain adaption. This model was able to make extrapolation predictions of the same sets to an error of 35Mpa for the same 6000 series aluminium alloys. Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London 2025-01-01T00:00:00Z