The question of how methods from the field of artificial intelligence can help improve the conventional frameworks for topology optimisation has received increasing attention over the last few years. Motivated by the capabilities of neural networks in image analysis, different model-variations aimed at obtaining iteration-free topology optimisation have been proposed with varying success. Other works focused on speed-up through replacing expensive optimisers and state solvers, or reducing the design-space have been attempted, but have not yet received the same attention. The portfolio of articles presenting different applications has as such become extensive, but few real breakthroughs have yet been celebrated. An overall trend in the literature is the strong faith in the “magic”of artificial intelligence and thus misunderstandings about the capabilities of such methods. The aim of this article is therefore to present a critical review of the current state of research in this field. To this end, an overview of the different model-applications is presented, and efforts are made to identify reasons for the overall lack of convincing success. A thorough analysis identifies and differentiates between problematic and promising aspects of existing models. The resulting findings are used to detail recommendations believed to encourage avenues of potential scientific progress for further research within the field. 

/pdf/on-the-use-of-artificial-neural-networks-in-topology-3hrzx4hk.pdf

On the use of artificial neural networks in topology optimisation

Additive manufacturing (AM) is the name given to a family of manufacturing processes where materials are joined to make parts from 3D modelling data, generally in a layer-upon-layer manner. AM is rapidly increasing in industrial adoption for the manufacture of end-use parts, which is therefore pushing for the maturation of design, process, and production techniques. Machine learning (ML) is a branch of artificial intelligence concerned with training programs to self-improve and has applications in a wide range of areas, such as computer vision, prediction, and information retrieval. Many of the problems facing AM can be categorised into one or more of these application areas. Studies have shown ML techniques to be effective in improving AM design, process, and production but there are limited industrial case studies to support further development of these techniques.

Machine Learning for Additive Manufacturing

State-of-the-Art Review of Machine Learning Applications in Additive Manufacturing; from Design to Manufacturing and Property Control

Abstract. The present study focuses on identifying the parameters from the Weather Research and Forecasting (WRF) model that strongly influence the simulation of tropical cyclones over the Bay of Bengal (BoB) region. Three global sensitivity analysis (SA) methods, namely, the Morris One-at-A-Time (MOAT), multivariate adaptive regression splines (MARS), and surrogate-based Sobol', are employed to identify the most sensitive parameters out of 24 tunable parameters corresponding to seven parameterization schemes of the WRF model. Ten tropical cyclones across different categories, such as cyclonic storms, severe cyclonic storms, and very severe cyclonic storms over BoB between 2011 and 2018, are selected in this study. The sensitivity scores of 24 parameters are evaluated for eight meteorological variables. The parameter sensitivity results are consistent across three SA methods for all the variables, and 8 out of the 24 parameters contribute 80 %–90 % to the overall sensitivity scores. It is found that the Sobol' method with Gaussian progress regression as a surrogate model can produce reliable sensitivity results when the available samples exceed 200. The parameters with which the model simulations have the least RMSE values when compared with the observations are considered the optimal parameters. Comparing observations and model simulations with the default and optimal parameters shows that simulations with the optimal set of parameters yield a 16.74 % improvement in the 10 m wind speed, 3.13 % in surface air temperature, 0.73 % in surface air pressure, and 9.18 % in precipitation simulations compared to the default set of parameters. 

/pdf/determining-the-sensitive-parameters-of-the-weather-research-1kfzoixs.pdf

Determining the sensitive parameters of the Weather Research and Forecasting (WRF) model for the simulation of tropical cyclones in the Bay of Bengal using global sensitivity analysis and machine learning

The prediction skill of a numerical model can be enhanced by calibrating the sensitive parameters that significantly influence the model forecast. The objective of the present study is to improve the prediction of surface wind speed and precipitation by calibrating the Weather Research and Forecasting (WRF) model parameters for the simulations of tropical cyclones over the Bay of Bengal region. Ten tropical cyclones across different intensity categories between 2011 and 2017 are selected for the calibration experiments. Eight sensitive model parameters are calibrated by minimizing the prediction error corresponding to 10m wind speed and precipitation, using a multiobjective adaptive surrogate model-based optimization (MO-ASMO) framework. The 10m wind speed and precipitation simulated by the default and calibrated parameter values across different aspects are compared. The results show that the calibrated parameters improved the prediction of 10m wind speed by 17.62% and precipitation by 8.20% compared to the default parameters. The effect of calibrated parameters on other model output variables, such as cyclone track and intensities, and 500 hPa wind fields, is investigated. Eight tropical cyclones across different categories between 2011 and 2018 are selected to corroborate the performance of the calibrated parameter values for other cyclone events. Finally, the robustness of the calibrated parameters across different boundary conditions and grid resolutions is also examined. These results will have significant implications for improving the predictability of tropical cyclone characteristics. This allows us to better plan the adaptation and mitigation strategies and thus help in preventing the adverse effects on society. 

/pdf/parameter-calibration-to-improve-the-prediction-of-tropical-1c8xpmke.pdf

Parameter Calibration to Improve the Prediction of Tropical Cyclones over the Bay of Bengal Using Machine Learning–Based Multiobjective Optimization

Topology optimization is a method to find the optimal material distribution of a structure by minimizing the objective function under the design and limit constraints. In this paper, we developed a deep learning-based machine learning algorithm to get the optimized structure for the given input conditions of a structure. We trained convolutional neural network (CNN)-based encoder–decoder architecture using the existing dataset as target images and input conditions modeled as input images. The target images are the optimized structures, developed using the MATLAB open-source topology optimization code, generated by varying the volume fraction from 5 to 95% with an increment of 5% and Poisson’s ratio varied from 0.01 to 0.49. The input conditions, i.e., the volume fraction and Poisson’s ratio are modeled as input images. In the present study, four types of input images and two encoder–decoder architectures are developed, and their performance is studied using identity mapping to obtain the optimized structure of a cantilever beam which is fixed at one end and a constant load is applied at the other end.

Topology optimization using convolutional neural network

The present study focuses on determining the best combination of microphysics (MP) and cumulus parameterization (CP) schemes for the simulation of Tropical Cyclones (TCs) in the Indian subcontinent region, using the Weather Research and Forecasting (WRF) model. From the available schemes, four CP schemes, namely Kain–Fritsch, Betts–Miller–Janjic, New Simplified Arakawa–Schubert, and Grell–Devenyi, and four MP schemes, namely WSM6, Purdue Lin, Thompson, and Morrison, are selected for the sensitivity study. Seven TCs are simulated using all combinations of the chosen physics schemes. The simulated tracks and intensities are compared against the India Meteorological Department (IMD) observations. The results show that the Kain–Fritsch scheme, in combination with all MP schemes, predicts the tracks best among all the available CP schemes, but the performance of microphysics schemes is indistinguishable. A further study is conducted using the Global Precipitation Measurement (GPM) radar data to identify the best MP scheme by comparing the reflectivities. An existing radar simulator is modified to calculate the simulated reflectivities from the WRF model output corresponding to the MP schemes. The simulated reflectivities are compared against the GPM radar reflectivities, and the results show that the Thompson scheme reproduces the reflectivities closest to observations. The performance of the best set of schemes obtained from this study is compared with a random set of schemes for cyclone Bulbul, and the best set of schemes outperformed the random schemes in every aspect.

A sensitivity study of WRF model microphysics and cumulus parameterization schemes for the simulation of tropical cyclones using GPM radar data

The prediction skill of a numerical model can be enhanced by calibrating the sensitive parameters that significantly influence the model forecast. The objective of the present study is to improve the prediction of surface wind speed and precipitation by calibrating the Weather Research and Forecasting (WRF) model parameters for the simulations of tropical cyclones over the Bay of Bengal region. Ten tropical cyclones across different intensity categories between 2011 and 2017 are selected for the calibration experiments. Eight sensitive model parameters are calibrated by minimizing the prediction error corresponding to 10m wind speed and precipitation, using a multiobjective adaptive surrogate model-based optimization (MO-ASMO) framework. The 10m wind speed and precipitation simulated by the default and calibrated parameter values across different aspects are compared. The results show that the calibrated parameters improved the prediction of 10m wind speed by 17.62% and precipitation by 8.20% compared to the default parameters. The effect of calibrated parameters on other model output variables, such as cyclone track and intensities, and 500 hPa wind fields, is investigated. Eight tropical cyclones across different categories between 2011 and 2018 are selected to corroborate the performance of the calibrated parameter values for other cyclone events. Finally, the robustness of the calibrated parameters across different boundary conditions and grid resolutions is also examined. These results will have significant implications for improving the predictability of tropical cyclone characteristics. This allows us to better plan the adaptation and mitigation strategies and thus help in preventing the adverse effects on society.

WRF model parameter calibration to improve the prediction of tropical cyclones over the Bay of Bengal using Machine Learning-based Multiobjective Optimization.

Diabetes is one of the common diseases in the world that cannot be permanently cured, but with proper medication one can lead a long and healthy life by curbing extreme complications. The skills and equipment required to identify the conditions take a longer time to provide an accurate result and are not an affordable means for all the income groups. In order to overcome this issue, an ML model is created and deployed in an application so it will be used by many in predicting the presence of the disease. The chapter focuses on detecting the presence of two major anomalies, namely diabetic retinopathy (DR) and glaucoma, which were caused due to diabetes. All the dataset used for the project is gathered from Kaggle and Messidor. Around six machine learning algorithms that fall under supervised learning techniques are executed. Among the many models, the random forest model has a high accuracy of 73% for DR prediction. Simultaneously, glaucoma detection is performed using different algorithms showing that Naive Bayes has the highest accuracy of 98%.

Automated Early Prediction of Anomalies Due to Diabetes Using Fundus Images

The present study focuses on identifying the parameters from the Weather Research and Forecasting (WRF) model that strongly influence the prediction of tropical cyclones over the Bay of Bengal (BoB) region. Three global sensitivity analysis (SA) methods namely the Morris One-at-A-Time (MOAT), Multivariate Adaptive Regression Splines (MARS), and surrogate-based Sobol' are employed to identify the most sensitive parameters out of 24 tunable parameters corresponding to seven parameterization schemes of the WRF model. Ten tropical cyclones across different categories, such as cyclonic storms, severe cyclonic storms, and very severe cyclonic storms over BoB between 2011 and 2018, are selected in this study. The sensitivity scores of 24 parameters are evaluated for eight meteorological variables. The parameter sensitivity results are consistent across three SA methods for all the variables, and 8 out of the 24 parameters contribute 80%-90% to the overall sensitivity scores. It is found that the Sobol' method with Gaussian progress regression as a surrogate model can produce reliable sensitivity results when the available samples exceed 200. The parameters with which the model simulations have the least RMSE values when compared with the observations are considered as the optimal parameters. Comparing observations and model simulations with the default and optimal parameters shows that predictions with the optimal set of parameters yield a 19.65% improvement in surface wind, 6.5% in surface temperature, and 13.2% in precipitation predictions, compared to the default set of parameters.

Determining the sensitive parameters of WRF model for the prediction of Tropical cyclones in Bay of Bengal using Global sensitivity analysis and Machine learning.

Balaji Srinivasan

Papers

Topology optimization using convolutional neural network

A sensitivity study of WRF model microphysics and cumulus parameterization schemes for the simulation of tropical cyclones using GPM radar data

WRF model parameter calibration to improve the prediction of tropical cyclones over the Bay of Bengal using Machine Learning-based Multiobjective Optimization.

Automated Early Prediction of Anomalies Due to Diabetes Using Fundus Images

Determining the sensitive parameters of WRF model for the prediction of Tropical cyclones in Bay of Bengal using Global sensitivity analysis and Machine learning.