In the demanding landscape of modern industrial manufacturing, achieving a balance between high-performance coating quality and operational cost-efficiency remains a persistent engineering challenge. A recent systematic study has introduced a robust framework for optimizing process parameters in high-velocity oxy-fuel (HVOF) thermal spraying, an essential technique for applying protective coatings to components subjected to harsh operating conditions. By integrating sophisticated mathematical modeling with three distinct multi-objective optimization algorithms, researchers have developed a comprehensive method that allows industry practitioners to navigate the complex trade-offs between critical process variables. This study addresses the necessity of precisely tailoring parameters—such as powder feed rate, stand-off distance, fuel-to-oxygen ratio, coating velocity, and total gas flow—to ensure that final coatings meet stringent performance criteria while simultaneously minimizing material waste and energy consumption.
Thermal spray technology, particularly the HVOF process, is integral to the steel industry, where it provides critical resistance to wear, corrosion, and high-temperature degradation, thereby extending the service life of equipment. The process relies on the combustion of fuel gases in a high-pressure chamber, generating supersonic gas velocities that accelerate powder particles onto a substrate to form dense, adherent coatings. Because the resulting coating properties—such as hardness, porosity, and roughness—are highly sensitive to input parameters and hardware configuration, the development of predictive models is essential. The research presented here utilizes generalized linear models (GLMs) based on gamma regression, which are particularly suited to handle variables that exhibit non-symmetrical, strictly positive distributions. These models, developed through a systematic design of experiments, provide the necessary predictive foundation for the subsequent optimization algorithms to identify Pareto-optimal solutions.
The study formally defines the multi-objective optimization problem within two essential Euclidean spaces: the decision variable space, where process parameters are manipulated, and the objective space, where the corresponding performance metrics are mapped. At the heart of this optimization effort is the search for Pareto-optimal solutions—trade-offs where no single objective can be improved without compromising another. The researchers evaluated three diverse optimization strategies: the Weighted Sum Method with Sequential Quadratic Programming (SQP), the Desirability approach, and the Non-dominated Sorting Genetic Algorithm II (NSGA-II). Each method was systematically assessed for its ability to guide the optimization process toward the Pareto-optimal front, highlighting the distinct advantages and constraints inherent in each approach within an industrial context.
The Weighted Sum Method is characterized by its simplicity and compatibility with standard single-objective optimization techniques, reformulating multi-objective problems by aggregating objectives into a single scalar function via user-defined weights. While it remains widely used, the study confirms that it often struggles to capture solutions in non-convex regions of the Pareto front, frequently defaulting to extreme boundary solutions. To address this, the researchers employed SQP, an iterative method effective for constrained nonlinear optimization, to solve these scalarized problems. Despite its efficiency, the limitation of the weighted sum approach in fully exploring complex objective spaces necessitated the use of more versatile, albeit computationally demanding, evolutionary strategies to ensure a thorough search of the available trade-off landscape.
In contrast, the desirability function approach provides a more intuitive framework, which is extensively utilized in materials science. This method transforms individual objectives into a dimensionless desirability value ranging from zero to one, allowing decision-makers to encode specific preferences before optimization begins. The overall desirability is then calculated as the geometric mean of these individual values, ensuring the final measure remains sensitive to poor performance in any single objective. Because this approach accommodates diverse quality characteristics and provides a flexible mechanism to prioritize specific criteria, it offers a pragmatic solution for industrial scenarios. However, the study notes that the accuracy of the optimization outcome is heavily dependent on the precision with which experts define the underlying desirability curves, requiring a high level of process-specific domain knowledge.
The Non-dominated Sorting Genetic Algorithm II (NSGA-II) serves as the third, and most computationally robust, optimization method explored in the study. NSGA-II integrates evolutionary principles with a sophisticated non-dominated sorting approach, allowing the algorithm to maintain a diverse set of solutions across the entire Pareto front. By incorporating an elite-preserving operator and a crowding distance metric, NSGA-II effectively prevents premature convergence and ensures that the final set of trade-offs is well-distributed. While the study acknowledges that NSGA-II requires significantly more computational power compared to gradient-based or desirability-based approaches, it concludes that this investment is secondary in industrial thermal spraying, where the preparation and application of large-scale components far outweigh the time spent on numerical optimization.
To demonstrate the practical application of this framework, the research team conducted three case studies focused on the optimization of tungsten carbide cobalt chrome (WC-10Co-4Cr) coatings, which are frequently used for their exceptional abrasion resistance. The first case study addressed the dual objectives of maximizing surface hardness and deposition efficiency. The second case study expanded the optimization to include a third objective: minimizing particle temperature, which is critical for protecting temperature-sensitive substrates from deformation or undesired phase transformations. Finally, the third case study focused on simultaneously minimizing coating porosity, surface roughness, and particle temperature. These scenarios provided a rigorous test of the models’ ability to navigate three-dimensional objective spaces and identify viable trade-off configurations under realistic operational constraints.
Following the theoretical identification of optimal parameter configurations, the researchers performed practical validation trials using an industrial HVOF spray booth. These trials served as a crucial step in assessing the reliability of the gamma regression models and the accuracy of the suggested optimization solutions. Samples were produced, mounted, and subjected to standard characterization techniques—including image-based porosity quantification, mobile hardness testing, and in-flight particle temperature measurement—to allow for a direct comparison with theoretical predictions. The results showed a high degree of agreement, with most objective deviations remaining below a few percentage points. Notably, while minor discrepancies were observed in porosity measurements, these were attributed to the methodological sensitivities of image analysis rather than flaws in the predictive models themselves.
The validation results provide strong evidence for the practical implementability of the proposed multi-objective optimization framework. The study effectively demonstrates that data-driven statistical modeling, when properly integrated with industry-recognized optimization algorithms, can provide a robust and systematic methodology for enhancing coating performance. By systematically accounting for material-specific boundaries and equipment-dependent limitations, the framework ensures that identified optimal points are not merely theoretical abstractions but are physically realizable parameters that can be directly applied to improve the durability and functionality of industrial components.
Despite the success of the methodology, the study identifies several areas for future improvement and expansion. A primary consideration is the need to generalize the findings to alternative coating materials beyond WC-10Co-4Cr, as well as the necessity of integrating real-time production data from steel manufacturing to move toward a more performance-oriented operational strategy. The evolving regulatory landscape, particularly regarding CO2 emissions and sustainability, necessitates coatings that can resist increasingly complex operational stresses. Consequently, ongoing research is exploring inductive transfer learning to infer unmeasured characteristics from target variables, as well as techniques to bridge the gap between small-scale experimental trials and large-scale industrial throughput.
In conclusion, this research establishes a data-driven framework that bridges the gap between complex mathematical optimization and the practical realities of industrial thermal spraying. By systematically evaluating the Weighted Sum, Desirability, and NSGA-II approaches, the study equips industry practitioners with a tailored toolset to balance conflicting performance goals. The integration of high-fidelity gamma regression models ensures that the optimization process is grounded in empirical reality, providing a structured pathway toward achieving operational excellence in surface engineering. As manufacturing demands continue to intensify, such quantitative methodologies will remain vital for developing next-generation coatings that are not only durable and efficient but also precisely tailored to the specific constraints of the environments they are designed to survive.
Subject of Research: Optimization of industrial thermal spraying parameters through multi-objective algorithms and predictive statistical modeling.
Article Title: A practical guide to optimizing industrial thermal spraying through comparative multi-objective optimization
Article References: Rannetbauer, W., Hubmer, S., Hambrock, C., & Ramlau, R. (2025). A practical guide to optimizing industrial thermal spraying through comparative multi-objective optimization. Journal of Mathematics in Industry, 15(1), Article 12. https://doi.org/10.1186/s13362-025-00177-w
Image Credits: AI Generated
DOI: 10.1186/s13362-025-00177-w
Keywords: Thermal spraying, HVOF, Multi-objective optimization, Surface engineering, Industrial manufacturing, Pareto optimality, NSGA-II, Gamma regression, Coating durability, Process parameter optimization, Wear resistance, Steel manufacturing, Computational modeling, Decision support systems, Predictive analytics, Statistical modeling, Performance metrics
Cite Scienmag News
Denise Maddox. (November 17, 2026). Optimizing Industrial Thermal Spray Coatings Through Multi-Objective Computational Frameworks. Scienmag. https://scienmag.com/optimizing-industrial-thermal-spray-coatings-through-multi-objective-computational-frameworks/
Denise Maddox. "Optimizing Industrial Thermal Spray Coatings Through Multi-Objective Computational Frameworks." Scienmag, 17 November 2026, https://scienmag.com/optimizing-industrial-thermal-spray-coatings-through-multi-objective-computational-frameworks/. Accessed 17 November 2026.
Denise Maddox. "Optimizing Industrial Thermal Spray Coatings Through Multi-Objective Computational Frameworks." Scienmag. November 17, 2026. https://scienmag.com/optimizing-industrial-thermal-spray-coatings-through-multi-objective-computational-frameworks/








