Two-phase Method Research Articles

A coupled discrete element material point method for fluid–solid–particle interactions with large deformations

Fluid–solid–particle systems such as water–soil–rock mixtures are very common in many natural processes and engineering applications. However, modelling these complex interactions is still very challenging due to their multi-scale nature. Here, we present a hybrid model that combines the advantages of the two-phase (solid and liquid) Material Point Method (MPM) on handling continuum materials with large deformations and the capability of the Discrete Element Method (DEM) on simulating the mechanical behaviours of rigid particles. Moreover, in our model, the DEM has the ability to deal with non-spherical particles. A new collision detection approach is presented in detail, and a unified DEM style contact force model is proposed to couple MPM with DEM by considering momentum exchanges between particles and continuum phases. The proposed model is validated by several numerical benchmarks, including (1) flows around a cylinder, (2) a block sliding on an inclined plane, (3) projectile impacts a granular medium, and (4) sphere impacts and penetrates into wet granular material. The results match well with analytical solutions and experimental observations, demonstrating this model’s capability to efficiently solve complex fluid–solid–particle interactions with large deformations. Finally, an example of column collapse is simulated to show future potential applications of the proposed method.

Drawing and Building Your Way to Understanding: The Importance of Constructing Multiple Representations When Learning Anatomy

IntroductionLearning a concept through engaging with multiple representations of it, has long been recognized as a powerful way to facilitate understanding. Yet, learning through the active construction of multiple representations of a particular concept has only recently begun to receive attention, and has yet to be fully explored in anatomy. Research shows that learning is more successful when learners are active in the construction of their own understanding. As such, drawing and building physical models in order to learn are two methods through which a student can actively construct multiple representations of a concept. The act of drawing and building models in learning anatomy are powerful ways to make the unseen seen and the complex simple.MethodsThis was a two‐phase project, repeated across three cohorts (2019, 2020 and 2021) of first year medical students. Phase 1 included development and implementation of a series of original drawing (‘Draw‐It’) and model building (‘Do‐It’) tasks for first year medical students at Warwick Medical School (3 cohorts; n=604). Activities covered all anatomical systems and regions of the body, with particular focus on threshold topics students traditionally find challenging. All Year 1 medical students participated in twenty‐five ‘Draw‐It’ / ‘Do‐It’ sessions, each 40‐minutes in length, integrated within their core anatomical curriculum spread across the academic year. Phase 2 involved a two‐phase sequential explanatory mixed methods study exploring how students engaged with and perceived the impact of the integrated drawing and model building tasks in the year 1 curriculum. This phase included an end of year quantitative survey (3 cohorts; n=598) and follow‐up semi‐structured one‐on‐one interviews (3 cohorts; n=56).ResultsQuantitative results from this study showed that 88% strongly agreed that the ‘Draw‐It’ and ‘Do‐It’ tasks helped them to simplify complex anatomy, 74% reported these tasks improved the way they visualised anatomy, and 86% reported that these tasks were key in showing them where their weak areas were. Qualitative results showed the majority of students felt the activities exposed their areas of misunderstanding, challenged them to visualise anatomy in a simpler way, increased their ability to remember complex subjects/concepts, and required them to draw upon material previously taught in the course, and re‐create it in their own way. Many students felt the sessions positively ‘took away the mystery’ surrounding complex subject areas, made drawing an ‘approachable skill’, required them ‘to think about the same anatomy in ways they never expected’, and encouraged them to think about the ‘why’ behind their anatomical learning.Conclusion and ImplicationsThis study has practical and theoretical benefit to the anatomical and medical education community, namely that active construction of understanding involves cognitive processes central to deep learning and encourages critical thinking. It offers a new understanding into the ways that drawing and model building activities can effectively be integrated throughout a curriculum, and shows how these dynamic learning tools can be used to inspire learners to take an active role in the construction of their own understanding through multiple representations of a topic.

Aerodynamic study of low Reynolds number airfoil and mini-unmanned aerial vehicle in simulated rain environment

PurposeRainfall is one of the main atmospheric conditions that significantly affect the aerodynamic performance of the low Reynolds number flights. In this paper, the adverse effects of rain on the aerodynamic performance of a two-dimensional (2D) airfoil with a chord-based low Reynolds number of 2 × 105 and the mini-unmanned aerial vehicle (UAV) for various flight conditions, i.e. 0°–40° at Mach number 0.04 were studied numerically. The purpose of this study is to explore the aerodynamic penalties that affect the liquid water content (LWC = 5.33) of the airfoil and UAV performance in rain under different flying conditions.Design/methodology/approachThe Eulerian–Lagrangian two-phase flow method is adopted to simulate the rain environment over an airfoil and mini-UAV aerodynamic performances. The Reynolds Averaged Navier–Stokes equations are considered to solve the time-averaged equations of motion for fluid flow.FindingsThe effect of rainfall on the airfoil and mini-UAV is studied numerically and validated experimentally. For 2D airfoil, the lift and drag coefficients for both numerical and experimental results show a very good correlation at Reynolds number 2 × 105. For three-dimensional (3D) mini-UAV, the lift and drag coefficients for both numerical and experimental results show a very good correlation at Mach number 0.04. The raindrops distribution around the airfoil, premature trailing edge separation, boundary-layer velocity profiles at five different chord positions (i.e. LE, 0.25c, 0.5c, 0.75c and 0.98c) on the upper surface of the airfoil, water film height and the location of rivulet formation on the upper surface of the airfoil are also presented.Originality/valueFor 2D airfoil, the recorded maximum variation of the coefficient of lift and lift-to-drag (L/D) ratio is observed to be 5.33% at an 8° and 10.53% at a 4° angle of attack (AOA) between numerical and experimental results under the influence of rainfall effect for LWC = 5.33. The L/D ratio percentage degradation is seen to be 61.9% at an AOA of 0°–2° for the rain environment. For 3D mini-UAV, the recorded maximum variation of the coefficient of lift and L/D ratio are observed to be 2.84% and 4.60% at a 30° stall AOA under the influence of rainfall effect for LWC = 5.33. The numerical results are impressively in agreement with the experimental results. UAV designers will benefit from the findings presented in this paper. This will be also helpful for training the pilots to control the airplanes in a rain environment.

Investigation on two-phase fluid mixture flow, heat transfer and entropy generation of a non-Newtonian water-CMC/CuO nanofluid inside a twisted tube with variable twist pitch: Numerical and evolutionary machine learning simulation

A 3-D numerical investigation is performed to evaluate the effect of variable twist pitch on the hydrothermal behavior and entropy generation features of non-Newtonian water-CMC/CuO nanofluid (NF) flow inside a twisted tube with a square cross-section. Three twisted tubes with a length of 500 mm, each of which has 3 twists, are considered. The first tube (Case I) has a constant twist pitch of 100 mm, while the twist pitch of Case II (150.0, 127.5, 100.8, 74.2, and 47.5 mm) and Case III (190, 144.6, 97.9, 51.3, and 6.2 mm) are variable. The simulations are performed using the two-phase mixture method considering different nanoparticle concentrations (ϕs) of 0–3% and Reynolds numbers (Res) of 600–1500. Based on the results, the highest and lowest overall hydrothermal performance was obtained for Case II and Case I, respectively. Moreover, the lowest ratios of thermal and frictional entropies of NF flow in the twisted tube to those of the plain tube were obtained for Case II. As another novelty of the current work, an evolutionary machine learning approach, namely, gene expression programming (GEP), was adopted to simulate the first law and second law performances of the NF in Case III.

Mechanical Behavior and Microstructural Study of Biocemented Sand under Various Treatment Methods

Biocementation mainly relies on the formation of calcium carbonate to bind soil particles. This paper investigates the effectiveness of biocementation in terms of treatment methods. The previously established two-phase methods are compared with newly developed one-phase-low-pH methods on their mechanical behavior and microstructure. The one-phase-low-pH methods present a higher urease fixation rate than two-phase methods, highlighting the cost-effectiveness of this method. A modified one-phase-low-pH using CH3COOH is also compared with HCl-based one-phase-low-pH method. The results show that the morphology and size of precipitated crystals have a great influence on strength development. An optimized treatment method based on the one-phase-low-pH method is also established, which is accomplished via the injection of 5 U/ml of bacterial culture together with 2 M of cementation solution during each treatment. After four times of treatments, a total of 7% cumulative calcium carbonate content can be obtained with an unconfined compressive strength of 2.15 MPa.

Diet and physical exercises for preschoolers with ADHD and their mothers: An intervention study

BackgroundAttention deficit hyperactivity disorder (ADHD) is a serious public health problem. Diet-focused approaches and physical exercise can be used to complement other ADHD management techniques. ObjectiveTo determine the prevalence of ADHD symptoms among preschoolers in nursery schools and to evaluate the educational interventions toward nutrition and physical exercise in mothers and their preschoolers with ADHD symptoms. Research methodologyA two-phase sampling method was employed. First, a cross-sectional survey was conducted to determine the prevalence of ADHD symptoms in four nursery schools (400 preschoolers aged between 3 and 6 years). Second, an intervention study (a quasi-experimental research design with one group completing the pre-test and the post-test) was performed on 36 preschoolers having ADHD symptoms and their mothers by using the educational intervention for mothers and photos and games about nutrition and physical exercise for the preschoolers with ADHD; mothers of four children out of the 40 refused to participate in the study. Data were analyzed using SPSS version 20. The paired t-test was used to determine significant differences between the groups. Differences were considered significant at P < 0.05. ResultsOf the 400 preschoolers, 10% had high ADHD symptoms. The mean score of mothers’ knowledge of nutrition and physical exercise improved after the implementation of the program (p = 0.01). In addition, preschoolers with ADHD enjoyed the session with photos and games (p = 0.01). Conclusions and ImplicationsEducational intervention significantly improved the knowledge of the mothers. Moreover, preschoolers with ADHD symptoms enjoyed the session with photos and games. This intervention appears to be feasible and promising for further investigation of its effects.

Finite-time stabilization of memristive neural networks via two-phase method

The finite-time stability and stabilization issues of memristive neural networks with time delays are studied via two-phase method, based on impulsive control and state feedback control. Different from classical control design, the control scheme here consists of two phases, i. e. the state can reach the inner of the unit ball exponentially from any initial condition in the first phase, and the then it can arrive at zero equilibrium from the inner of the unit ball in a finite-time. By using different Lyapunov function in two phases, both the finite-time stability criteria based on algebra inequality and the settling time dependent on the initial condition are obtained simultaneously. Compare with the finite-time control strategy in the literature, the finite-time stability and finite-time controller for stabilization of memristive neural networks are used in the second phase only, rather than the whole control process. In addition, the pinning control based on our novel control scheme is designed to obtain the finite-time stabilization of memristive neural networks.

Designing and evaluating falls prevention education with residents and staff in aged care homes: a feasibility study

PurposeThe purpose of this study was to co-design a falls prevention education programme with aged care home residents and staff and evaluate its feasibility. The intention of providing the education programme was to assist residents to stay safe and mobile whilst reducing their risk of falling.Design/methodology/approachA two-phase mixed methods participatory design using a resident (n = 6) and care staff (n = 5) consumer engagement panel, pre- and post-programme resident (n = 35) survey and semi-structured care staff interviews (n = 8) was undertaken in two countries.FindingsA poster, brochure, video and staff education guide featuring 12 safety messages depicting fall prevention behaviours were co-designed. Residents, supported by staff, perceived the falls prevention education programme as enjoyable and informative, but there were no significant differences in capability, opportunity or motivation. However, several residents were observed enacting fall prevention behaviours such as “If I feel unwell, I'll ring the bell” and waiting for staff assistance. Challenges to programme demand, acceptability and implementation which may have impacted residents' exposure and engagement with the programme were identified, along with recommendations to improve feasibility.Practical implicationsWhen developing falls prevention education programmes partnering with residents and staff, providing choices to meet personal and aesthetic preferences along with frequent, shorter duration learning opportunities are important for translating education messages into actions.Originality/valueThe use of bespoke resources, novel rhymes, positive messages emphasising safety and co-designing with residents themselves was a welcomed point of programme difference.

Image identification for two-phase flow patterns based on CNN algorithms

Flow patterns are essential and useful to model the interfacial structures and heat transfer in gas-liquid two-phase flow. However, the current two-phase flow patterns classification methods mostly depend on direct visual observation. This study adopted a new flow pattern classification method based on convolutional neural network (CNN) algorithms to achieve an automatic and objective identification of two-phase flow patterns. A database of 696 test conditions, including 105642 condensing flow pattern images of methane and tetrafluoromethane in a horizontal circular tube, is collected as the input of the data-driven algorithms. After 80% of image data is fed to train and fit the parameters in the algorithms, the trained models with acceptable universality are obtained to identify five flow patterns: annular flow, bubbly flow, churn flow, slug flow and stratified flow. Compared with the manual classification, the proposed method can accurately predict two-phase flow patterns with a prediction accuracy of more than 90.63% and 91.45% for the test dataset and the entire database, respectively. The average accuracy for predicting all data points in the database is more than 97.56%. The results showed that using images as input, CNN algorithms can provide objective prediction with satisfactory accuracy and universality for two-phase flow pattern identification.

Process quality recheck for Gamma quality characteristic from supplier products: a case study on radio-frequency power

Quality fraud seriously damages the right-to-known of customers. Therefore, it is necessary for customers to recheck the process quality level declared by suppliers, which is often measured by process capability indices (PCIs). However, there exist two practical problems when rechecking PCIs based on a quality characteristic (QC). First, QC data often do not follow normal distributions. Second, to enhance the market competitiveness, the supplier usually conducts a full inspection and eliminates the non-conforming items before selling them, which causes the QC data to be truncated. To overcome these problems, motivated by the radio-frequency (RF) power output data, this paper proposes a two-phase process capability recheck method, including data-filling and generalised p-value (GE-P) test. The novel data-filling method integrates the quantile-filling (QA) and bias-correction closed-form (BC) estimator for Gamma distributions, and simulation results show it performs better than the traditional methods on unbiasedness and consistency. Based on the filling pseudo-complete data, GE-P is adopted to complete the recheck by testing whether the process capability reaches the supplier declared level. Numerical analysis indicates it performs well on two types of errors. Finally, a real case study on the motivating example is presented to verify the effectiveness of the proposed method.

Hydro-mechanical behaviour of soils during water-soil gushing in shield tunnels using MPM

Water-soil gushing has become the main risk affecting the safety of shield tunnels in water-rich sandy layers. In this paper, a two-phase single-point material point method (MPM) is used to simulate the evolution of gushing in shield tunnels and the hydro-mechanical responses of the surrounding soils. The method is validated by well-documented experiments. The influence of gushing on the surrounding soil can be divided into three zones: flow zone, disturbed zone, and stationary zone. Based on the development laws of the flow zone, the gushing process can be divided into three stages: the initial developing stage, the expansive developing stage, and the stable developing stage. Parametric analysis is carried out to investigate the effect of the gushing location, the buried depth of the tunnel, and the soil parameters. The lower the gushing location, the smaller the development rate of the gushing, but the greater the variation in soil stress. The shallower the tunnel depth for the gushing, the smaller the impact on deep layers, but the greater the influence on ground settlement. Increasing the friction angle can effectively reduce the area and the development rate of the flow zone.

Factors That Affect Methane Yield Using Raw Olive Alperujo (Unhydrolyzed) as Substrate in BMP Assays

The olive alperujo (OA) corresponds to the solid waste generated in the olive oil extraction process using the two-phase centrifugation method. OA is produced in large quantities (800 kg OA/ton olives processed) and is characterized by its high moisture content, organic matter, and low pH. In Chile, the olive oil industry is recent, and one of its main challenges is to be able to manage OA to reduce the impact caused by its disposal. In this sense, its valorization as biogas by means of anaerobic digestion is an economically attractive option. For this, it is previously necessary to evaluate the biomethane potential (BMP) of the raw OA using batch assays. This study was focused on evaluating the factors that most affect the methane yield (MY) when using OA as substrate in BMP tests. First, a sweep analysis (Plackett–Burman) was applied to determine those factors that, according to the literature, would have an influence on the BMP tests. Among the factors studied, the most significant were preincubation, OA concentration, and agitation level. Subsequently, a 23 factorial experimental design was applied to evaluate the effect of these factors on MY at different levels. Results show that the OA concentration was the most significant factor affecting MY.

Optimal Pressure Management in Water Distribution Systems: Efficiency Indexes for Volumetric Cost Performance, Consumption and Linear Leakage Measurements

Water is a finite resource and should be given the attention it deserves to reduce its depletion through leakages in pipe systems. The authors implemented pressure management strategies linked to fixed and variable discharge (FAVAD), the burst and background estimate (BABE), and orifice principle methodologies to analyze a two-phased comparative method for applying optimal pressure management and its efficiency indexes in measuring volumetric cost performance, consumption, leakage flowrate, linear leakage reduction, infrastructural leakage and leakage cost indices. Using time-modulated smart control pressure reducing valve (PRV) simulation processes, the authors selected Alexandra Township in Johannesburg, South Africa as a case study. The results showed a reduction in head pressure, a reduction in the system input volume (SIV) from 26,272,579 m3 to 21,915,943 m3 and a reduction in minimum night flow (MNF) from 14.01% to 12.50%. The annual estimated nodal system output (NSO) was reduced from 14,774.62 m3 to 12,787.85 m3. The monthly average linear system repairs were reduced from 246 to 177, while the efficiency index percentages of leakage frequency/km/pressure were reduced from 8.31% to 5.98%. At a unit cost of $3.18/m3, the cost of leakages declined from $4,009,315.54 to $2,862,053.10 per month, while average household consumption (AMC) reduced from 36.33 m3 to 24.56 m3. Finally, the linear reduction value R2 for the percentage of the total leakage flowrate (TLFR)/SIV declined from 0.58 to 0.5, whereas the infrastructure leakage ratio (ILI) increased from 4 to 4.3. The results fully demonstrated that optimal pressure management is an alternative way to simulate, estimate, quantify and understand where and how water is lost in a distribution system. The authors propose that the implementation of proactive leakage management and domestic background leakage repair could further assist in reducing the frequency and cost of water leakages.

Experimental investigation on the heat removal capacity of secondary side passive residual heat removal system for an integrated reactor

An experimental facility named Intermediate Heat Exchange Loop (IHEL) is built to simulate the secondary side passive residual heat removal system. The IHEL mainly includes a helically coiled Once-Through Steam Generator (OTSG), a condenser, a cooling water tank, and associated piping and valves. The steam is produced by the water-to-water heat exchange method in the SG secondary side. The heat from the SG primary side is carried to the cooling water tank by the IHEL through a two-phase natural circulation method. The natural circulation characteristics of the IHEL under the low pressure (nearly atmospheric pressure) and low water inventory conditions are investigated. The result shows that the heat removal capacity of the IHEL is good at these conditions, and it is determined by the parameters of the SG primary side and IHEL together. In the present test conditions, for the parameters of the SG primary side, the heat removal rate will increase with the increase of mass flow rate or heat source temperature. For the parameters of the IHEL, the heat removal rate is not sensitive to the initial non-condensable gas content. The heat removal rate is not sensitive to the resistance level when the heat source temperature is low, while it will decrease with the increase of resistance level when the heat source temperature is high. A very meaningful conclusion is that the natural circulation has been established successfully in the very lower water inventory condition (L = 14.8% in the present study), and the heat removal rate almost remains constant with the increase of the water inventory, which indicates the IHEL is robust and is beneficial for the system design.

Two-phase ESO and comprehensive learning PSO method for structural optimization with discrete steel sections

This paper proposes the novel two-phase metaheuristic method that combines the distinctive features provided by evolutionary structural optimization (ESO) and comprehensive learning particle swarm optimization (CLPSO) for the design of steel structures with standard sections. The approach overcomes the challenges associated with local optima pitfalls in processing the mixed integer nonlinear programming problem, and hence determines the accurate solutions at modest computing efforts. In essence, the first-phase ESO design incorporates the constraint relaxations to fast eliminate the ineffective (infeasible) sections from initial discrete variable domains. This advantageously provides a significant reduction in the search spaces performed in the second-phase CLPSO scheme. The comprehensive learning technique enhances a standard particle swarm optimization algorithm by enabling the cooperative responses among swarm populations. The learning probability function defines the comprehensive cross-positions between the sets of best particles constructed from the reducing sets of steel sections specified in the preliminary ESO phase leading to the computation of optimal solutions. Various benchmarks (subjected to infinite design combinations) illustrate robustness and accuracy of the proposed design method (viz., generating the smaller number of particles) as compared to conventional metaheuristic algorithms.

A Two-Phase Evolutionary Method to Train RBF Networks

This article proposes a two-phase hybrid method to train RBF neural networks for classification and regression problems. During the first phase, a range for the critical parameters of the RBF network is estimated and in the second phase a genetic algorithm is incorporated to locate the best RBF neural network for the underlying problem. The method is compared against other training methods of RBF neural networks on a wide series of classification and regression problems from the relevant literature and the results are reported.

An insight into core flooding experiment via NMR imaging and numerical simulation

Traditional core flooding experiments can only be used post breakthrough while what happens in the core prior to this time is vital to understand multiphase flow phenomenon for more successful EOR operations. We can overcome this obstacle through a visualized fluid displacement scheme. This can ultimately provide us with a reliable relative permeability curve that can lead to a more accurate reservoir simulation outcome in the field scale. In this study, NMR imaging is employed in a water flood experiment in conjunction with two separate numerical two-phase flow simulation methods (FDM and FEM), to reproduce experimental data. Using the Brooks-Corey equation, random pore size distribution indices (λ) are selected to generate relative permeability curves. Moreover, simulations are performed with FDM, and oil displacement efficiency, saturation maps, and saturation profiles are generated and compared to the experimental results. Next, FEM was employed in COMSOL for further validation and FDM results were found in agreement with the experiments. This way, an appropriate relative permeability curve was generated and assigned to the sample. Results suggest that λ of 0.2 generated the best numerical results with an MSE value of 0.009 in oil displacement efficiency curves, comparable to the experiments. Collectively, integration of imaging techniques with routine experimental fluid displacement procedures presented a detailed insight into complicated nature of multiphase flow phenomena in geomaterials.

Better understanding care transitions of adults with complex health and social care needs: a study protocol

BackgroundAdults with chronic conditions who also suffer from mental health comorbidities and/or social vulnerability require services from many providers across different sectors. They may have complex health and social care needs and experience poorer health indicators and high mortality rates while generating considerable costs to the health and social services system. In response, the literature has stressed the need for a collaborative approach amongst providers to facilitate the care transition process. A better understanding of care transitions is the next step towards the improvement of integrated care models. The aim of the study is to better understand care transitions of adults with complex health and social care needs across community, primary care, and hospital settings, combining the experiences of patients and their families, providers, and health managers.Methods/designWe will conduct a two-phase mixed methods multiple case study (quantitative and qualitative). We will work with six cases in three Canadian provinces, each case being the actual care transitions across community, primary care, and hospital settings. Adult patients with complex needs will be identified by having visited the emergency department at least three times over the previous 12 months. To ensure they have complex needs, they will be invited to complete INTERMED Self-Assessment and invited to enroll if positive. For the quantitative phase, data will be obtained through questionnaires and multi-level regression analyses will be conducted. For the qualitative phase, semi-structured interviews and focus groups will be conducted with patients, family members, care providers, and managers, and thematic analysis will be performed. Quantitative and qualitative results will be compared and then merged.DiscussionThis study is one of the first to examine care transitions of adults with complex needs by adopting a comprehensive vision of care transitions and bringing together the experiences of patients and family members, providers, and health managers. By using an integrated knowledge translation approach with key knowledge users, the study’s findings have the potential to inform the optimization of integrated care, to positively impact the health of adults with complex needs, and reduce the economic burden to the health and social care systems.

Removing the impulse noise from images based on fuzzy cellular automata by using a two-phase innovative method

In this paper, a novel method is proposed to restore digital images corrupted by impulse noise. The proposed method in this paper consists of two phases for detecting impulse noise and image restoration. Corrupted pixels in the first phase of the proposed method are identified in two steps. In the first step, the corrupted pixels are identified by minimum values and the average Moore neighborhood pixels of the central pixel. In the second step of this phase, the pixels identified in the previous step as an uncorrupted pixel, are examined whether they are uncorrupted or corrupted once again for further accuracy by a new algorithm. The new algorithm uses cellular automata to identify corrupted pixels by measuring the harmonic and arithmetic means values of four different states of the position of five pixels in the neighborhood of the central pixel. In the second phase of the proposed method, the corrupted pixels are restored using the cosine similarity of the maximization of four different positions of the five pixels neighborhood to the central pixel using fuzzy cellular. The proposed algorithm is evaluated using image enhancement factor (IEF), Structural Similarity Index (SSIM), and Peak Signal to Noise Ratio (PSNR). The proposed method has been tested on 580 images at 15–90% noise density and the qualitative and quantitative results show that the proposed filter is robust enough at different noise levels and meaningful image details such as edges compared to other methods are well maintained.

Bayesian calibration of multi-level model with unobservable distributed response and application to miter gates

Bayesian calibration plays a vital role in improving the validity of computational models’ predictive power. However, the presence of unobservable distributed responses and uncertain model parameters in multi-level models poses challenges to Bayesian calibration, due to the lack of direct observations and the difficulty in identifying the hidden and distributed model discrepancy under uncertainty. This paper proposes a Bayesian calibration framework for multi-level simulation models to calibrate an unobservable distributed model using measurements of an observable model. In the proposed framework, the distributed model discrepancy of an unobservable model with distributed response is first represented as a series of orthogonal polynomials, with the polynomial coefficients modelled by surrogate models with unknown hyper-parameters. A two-phase machine learning method is then developed to construct surrogate models of the polynomial coefficients based on measurements of an observable model. The constructed model discrepancy is finally used to update the uncertain model parameters by following a modularized Bayesian calibration scheme. The developed framework is applied to the joint Bayesian calibration of the uncertain gap length and unobservable and distributed boundary condition model for a miter gate problem. Results of the miter gate application demonstrate the efficacy of the proposed framework.