Accelerating Global Sensitivity Analysis via Supervised Machine Learning Tools: Case Studies for Mineral Processing Models

Social housing in tropical regions faces critical thermal comfort challenges that will intensify under future climate change, yet current design practices lack systematic frameworks for evaluating long-term performance across multiple climate scenarios. This study assesses the thermal performance of social housing in southeastern Mexico using energy simulation, supervised machine learning, and global sensitivity analysis. Two housing typologies (single-story and two-story) were modeled across four cities (Mérida, Campeche, Cancún, and Tuxtla Gutiérrez) under climate change scenarios (RCP 2.6, 4.5, and 8.5) for 2050 and 2100. Various machine learning models were trained to predict comfort temperature and cooling degree days. Regression Trees demonstrated superior performance, with R2 values exceeding 0.98 for both thermal comfort indicators, achieving RMSE values of 0.0095 °C for comfort temperature and 0.2613 °C for cooling degree days. Global sensitivity analysis using the PAWN method revealed that ambient temperature was the most influential variable, accounting for 45–49% of the total sensitivity, followed by solar radiation (17–22%) and relative humidity (10–12%), while building-specific parameters had modest impacts (0.6–3.8%). Geographic variations were significant, with Mérida and Campeche showing higher cooling demands than Cancún and Tuxtla Gutiérrez. Future climate projections indicate substantial increases in cooling requirements by 2100, with CDD values expected to increase by approximately 40–50% under the RCP 8.5 scenario compared to current conditions. This research presents a computational framework for assessing thermal comfort in social housing, providing evidence-based insights for climate-adaptive building strategies in tropical regions.

Abstract. Integrated hydrological modeling of domains with complex subsurface features requires many highly uncertain parameters. Performing a global uncertainty analysis using an ensemble of model runs can help bring clarity as to which of these parameters really influence system behavior and for which high parameter uncertainty does not result in similarly high uncertainty of model predictions. However, already creating a sufficiently large ensemble of model simulation for the global sensitivity analysis can be challenging, as many combinations of model parameters can lead to unrealistic model behavior. In this work we use the method of active subspaces to perform a global sensitivity analysis. While building up the ensemble, we use the already-existing ensemble members to construct low-order meta-models based on the first two active-subspace dimensions. The meta-models are used to pre-determine whether a random parameter combination in the stochastic sampling is likely to result in unrealistic behavior so that such a parameter combination is excluded without running the computationally expensive full model. An important reason for choosing the active-subspace method is that both the activity score of the global sensitivity analysis and the meta-models can easily be understood and visualized. We test the approach on a subsurface-flow model including uncertain hydraulic parameters, uncertain boundary conditions and uncertain geological structure. We show that sufficiently detailed active subspaces exist for most observations of interest. The pre-selection by the meta-model significantly reduces the number of full-model runs that must be rejected due to unrealistic behavior. An essential but difficult part in active-subspace sampling using complex models is approximating the gradient of the simulated observation with respect to all parameters. We show that this can effectively and meaningfully be done with second-order polynomials.

This paper presents an application of the Global Sensitivity Analysis (GSA) method in evaluating performance of impedance measurement algorithms implemented in modern Intelligent Electronic Devices (IEDs) when affected by uncertainty. To be specific, the GSA is applied in testing an algorithm for distance protection of a single transmission line in a system with two lines on the same towers. Factors contributing to uncertainty are discussed in the paper. The GSA method decomposes variance of the estimation error according to the factors and their interactions, and in this way identifies the factors that have the highest impact on the estimation error. Computation of the GSA sensitivity indices relies on the Quasi-Monte Carlo (QMC) technique with Sobol's quasi-random sampling. This is time-consuming for large number of factors therefore, fast pre-screening process based on the Morris method is implemented to remove non-influential parameters and reduce the factor space before running the QMC. The methodology is implemented using fault simulation software DIgSILENT PowerFactory to calculate fault impedance as a function of uncertain parameters (i.e. factors) as well as voltage and current signals that are inputs to an IED under test. In this simulation, factor space is sampled using Sobol's quasi-random sequence generated with the standard GSA software tool called SIMLAB. All simulations of the transmission line faults for various parameter values and calculation of the performance indices (i.e. sensitivities) is automated using the program developed in DIgSILENT Programming Language (DPL). This testing technique is applied to study the fault impedance measurement implemented in SEL-421. (6 pages)

Variance-based global sensitivity analysis (GSA) can provide a wealth of information when applied to complex models. A well-known Achilles' heel of this approach is its computational cost, which often renders it unfeasible in practice. An appealing alternative is to instead analyze the sensitivity of a surrogate model with the goal of lowering computational costs while maintaining sufficient accuracy. Should a surrogate be "simple" enough to be amenable to the analytical calculations of its Sobol' indices, the cost of GSA is essentially reduced to the construction of the surrogate.We propose a new class of sparse-weight extreme learning machines (ELMs), which, when considered as surrogates in the context of GSA, admit analytical formulas for their Sobol' indices and, unlike the standard ELMs, yield accurate approximations of these indices. The effectiveness of this approach is illustrated through both traditional benchmarks in the field and on a chemical reaction network.

Multiphase systems are important in minerals processing, and usually include solid–solid and solid–fluid systems, such as in wet grinding, flotation, dewatering, and magnetic separation, among several other unit operations. In this paper, the current trends in the process system engineering tasks of modeling, design, and optimization in multiphase systems, are analyzed. Different scales of size and time are included, and therefore, the analysis includes modeling at the molecular level (molecular dynamic modeling) and unit operation level (e.g., computational fluid dynamic, CFD), and the application of optimization for the design of a plant. New strategies for the modeling, design, and optimization of multiphase systems are also included, with a strong focus on the application of artificial intelligence (AI) and the combination of experimentation and modeling with response surface methodology (RSM). The integration of different modeling techniques such as CFD with discrete element simulation (DEM) and response surface methodology (RSM) with artificial neural networks (ANN) is included. The paper finishes with tools to study the uncertainty, both epistemic and stochastic, based on uncertainty and global sensitivity analyses, which is present in all mineral processing operations. It is shown that all of these areas are very active and can help in the understanding, operation, design, and optimization of mineral processing that involves multiphase systems. Future needs, such as meso-scale modeling, are highlighted.

Identifying key sources of uncertainty in the modelling of greenhouse gas emissions from wastewater treatment

Advances in our ability to model complex environmental systems are currently driven by at least four needs: (1) the need for the inclusion of uncertainty in monitoring, modelling and decision-making; (2) the need to provide environmental predictions everywhere; (3) the need to predict the impacts of environmental change; and (4) the need to adaptively evolve observation networks to better resolve environmental systems and embrace sensing innovations. Satisfying these needs will require improved theory, improved models and improved frameworks for making and evaluating predictions. All of these improvements should result in the long-term evolution and improvement of observation systems. In the context of this paper we discuss current bottlenecks and opportunities for advancing environmental modelling with and without local observations of system response. More realistic representations of real-world thresholds, nonlinearities and feedbacks motivates the use of more complex models as well as the consequent need for more rigorous evaluations of model performance. In the case of gauged systems, we find that global sensitivity analysis provides a widely underused tool for evaluating models' assumptions and estimating the information content of data. In the case of ungauged systems, including the modelling of environmental change impacts, we propose that the definition of constraints on the expected system response provides a promising way forward. Examples of our own work are included to support the conclusions of this discussion paper. Overall, we conclude that an important bottleneck currently limiting environmental predictions lies in how our model evaluation and identification approaches are extracting, using and evolving the information available for environmental systems at the watershed scale.

Aimed at the structural parameters in the new optical Y-branch, this paper uses the Morris screening method and Sobol method in global sensitivity analysis to analyze the sensitivity of each parameter when the input optical field introduces an offset. The sensitivity parameters of the optical Y-branch are selected, and global sensitivity analysis of the sensitivity parameters is performed. The results of sensitivity analysis improve the parameter optimization process of the optical Y-branch. Finally, the optical Y-branch is optimized to obtain a small insertion loss, good uniformity, low wavelength-dependent loss, and polarization-dependent loss.

The movement of currents in coastal areas is greatly influenced by fluctuations in water elevation. The dynamics of the sea and its relationship with the surrounding environment make the approach to current patterns and sediment distribution complex. In line with technological advances, this approach can be carried out using numerical models of hydrodynamic equations. The quality of hydrodynamic modelling depends on the quality of the variables used. However, the availability of data that can be used spatially and/or temporally is very limited. Therefore, the study will be conducted by considering the contribution of wind, discharge, tidal, and sediment concentration variables to sediment accumulation values and current velocity. Sensitivity analysis was performed on two sample points with different characteristics in two seasons. The method used was the GLS ARMA regression equation estimation approach in defining the modelling equation, and sensitivity analysis was performed using the Global Sensitivity Analysis (GSA) Sobol method. The regression equation shows a significant p-value in the GLS ARMA model (0,1). Testing was conducted on the equation model using the Sobol method to identify sensitive variables. In current modelling, the discharge variable has the largest direct contribution to current variation. In sedimentation modelling, the concentration variable has the largest direct contribution to sediment variation. In addition to direct effects, the relationship between wind and discharge contributed -0.0046 and 0.0023 in the West and East seasons, respectively, to the sedimentation model results. In the flow modelling, the relationship between wind and discharge contributed -0.0040 and 0.0047 in the West and East seasons, respectively. The modelling process was repeated by providing two change scenarios based on directly sensitive variables and compared using the RRMSE method. In flow modelling, the change in results was 66% in each change scenario between the average data of the change model and the actual model. Meanwhile, in sediment modelling, the change was 37% and 334% in each change scenario between the average data of the change model and the actual model.

To enhance the accuracy of advanced constitutive models for soft natural clays, several parameters are necessary, resulting in complexity of numerical modelling. However, the detailed effects of these parameters are not rigorously quantified towards their constitutive relationships. Thus, the aim of the paper is to assess such advanced models in order to determine the most significant parameters over the time series data. Two methods for Global Sensitivity Analysis (GSA), i.e. Experimental Design and the Sobol method, were used and benchmarked to assess the model predictions on a discretised domain in time and space. A rate-dependent Finite Element model using Creep-SCLAY1S and a hydro-mechanically coupled formulation for consolidation was used to study a Constant Rate of Strain (CRS) test. The value of GSA approaches adopted herein was to investigate the model predictions both in the temporal and spatial domain. The temporal analyses indicate three sets of significant model parameters in different portions of the CRS compression curve. Furthermore, the non-stationary nature of sensitivity results is exposed, identifying the parameters that lead to unique solutions for the CRS loading path. The FE implementation enabled the quantification of the most sensitive model parameters in the spatial domain. The spatial results that are governed by the rate-dependent processes in the soil (i.e. consolidation and creep) illustrated that Experimental Design was capable of providing sensitivity maps with satisfactory accuracy similar to the Sobol method. Experimental Design was found to be the most efficient method, concerning execution time and storage costs, to assess rate-dependent problems in Geotechnics.

An LS-SVM classifier based methodology for avoiding unwanted responses in processes under uncertainties

To understand the change trend of Lake Taihu’s water quality indexes in advance and prevent the occurrence of pollution incidents such as blue algae, the ammonia nitrogen index forecasted model was studied by taking the water quality monitoring data of Nanquan in Wuxi City as an example. The chromaticity, odor and taste, pH, nitrite nitrogen, chloride, dissolved oxygen, and algae were determined as input variables by implementing linear regression analysis through SPSS software to establish the Extreme Learning Machine (ELM) model for forecasting ammonia nitrogen index. Aiming at the problem of ELM randomly generating weights and thresholds resulting in low forecast accuracy, the Particle Swarm Optimization (PSO) algorithm was adopted to optimize ELM parameters. In order to solve the situation that the PSO algorithm is easy to fall into local optimum, the Adaboost algorithm was introduced to train weak ELM predictors, which were later combined into a strong predictor for the forecast. The results represent that the Adaboost-PSO-ELM model’s forecast coefficient of determination (\({R}^{2}\)) and root mean square error (\(RMSE\)) for ammonia nitrogen are 0.955 mg/L and 0.023 mg/L, respectively. Compared with the two models of ELM and PSO-ELM, the forecast accuracy is significantly improved. Therefore, the ELM model based on Adaboost and PSO algorithms is feasible for forecasting water quality index of ammonia nitrogen in Lake Taihu, providing new ideas for accurate forecast in water quality aspect.KeywordsLake TaihuAmmonia nitrogenAdaboost-PSO-ELM

The long-term integrity of the bentonite buffer is of significant interest in the performance assessment (PA) of geological nuclear waste disposal. This study aims at understanding how the initial bentonite chemical composition and other geochemical parameters affect long-term chemical properties within the buffer, which will subsequently affect the transport. Using coupled thermal–hydrological–chemical (THC) models for migration of U(VI) in a generic repository, we performed a global sensitivity analysis (GSA) to identify the influence of each parameter on the temporal evolution of a spatially averaged distribution coefficient for the entire buffer. Such an analysis can be used in a repository-scale PA. In this work, we used the toughreact software to model coupled THC processes in a generic clay repository with bentonite buffer. In this model, U(VI) is released from a canister via schoepite dissolution, which is assumed to occur 1000 yr after closure. U(VI) migrates through the bentonite buffer affected by two-site protolysis nonelectrostatic surface complexation and cation exchange (2 SPNE SC/CE). GSA results showed that adsorption density on smectite, pH, volume fractions of smectite, calcite, and Ca+2 aqueous concentration all play a significant role in U(VI) transport, since roughly 80% of adsorbed U(VI) is absorbed by smectite, and Ca+2 affects the aqueous complexation with U(VI). This work demonstrates the complex process models' potential usefulness that can be transferred to the PA model. It also provides information needed to proceed with the development of a reduced-order model, which has the potential to optimize repository designs, site characterization, and performance confirmation.

The dynamic modulus (∣E*∣) of hot-mix asphalt mixes is one of the most time-consuming and labour-intensive material metrics to evaluate in the laboratory. This study introduces a novel paradigm for assessing the ∣E*∣'s most influential elements by employing widely accepted literature models. Witczak 1-37A, Witczak 1-40D, and Modified Hirsch Models are selected for analysing the asphalt dynamic modulus. First, a thorough laboratory database of Arizona State University is used to account for all major input factors, such as mixture gradation, binder qualities, volumetric properties, and testing conditions parameters, during models' validation. Second, each model's performance is evaluated using standard measures to build confidence levels in the subsequent analysis stage. Finally, with the aid of Latin Hypercube Simulation, a comprehensive global sensitivity analysis (GSA) is performed. Three unique GSA approaches are used; namely, elementary effects, variance-based, and PAWN methods, to highlight the effect of each input variable on the magnitude of ∣E*∣. Different GSA tools are strongly recommended since there is no analytical tool for validating the findings with the complex formulations of the selected mathematical models. The GSA demonstrates that the voids ratio in total mix, binder shear modulus, viscosity, phase angle, and binder quantity are the most significant factors.

Robustness of Extreme Learning Machine in the prediction of hydrological flow series