Fast prediction of distances between synthetic routes with deep learning

We expand the recent work on clustering of synthetic routes and train a deep learning model to predict the distances between arbitrary routes. The model is based on an long short-term memory (LSTM) representation of a synthetic route and is trained as a twin network to reproduce the tree edit distance (TED) between two routes. The ML approach is approximately two orders of magnitude faster than the TED approach and enables clustering many more routes from a retrosynthesis route prediction. The clusters have a high degree of similarity to the clusters given by the TED-based approach and are accordingly intuitive and explainable. We provide the developed model as open-source.

Research on Chest Disease Recognition Based on Deep Hierarchical Learning Algorithm

Chest X-ray has become one of the most common ways in diagnostic radiology exams, and this technology assists expert radiologists with finding the patients at potential risk of cardiopathy and lung diseases. However, it is still a challenge for expert radiologists to assess thousands of cases in a short period so that deep learning methods are introduced to tackle this problem. Since the diseases have correlations with each other and have hierarchical features, the traditional classification scheme could not achieve a good performance. In order to extract the correlation features among the diseases, some GCN-based models are introduced to combine the features extracted from the images to make prediction. This scheme can work well with the high quality of image features, so backbone with high computation cost plays a vital role in this scheme. However, a fast prediction in diagnostic radiology is also needed especially in case of emergency or region with low computation facilities, so we proposed an efficient convolutional neural network with GCN, which is named SGGCN, to meet the need of efficient computation and considerable accuracy. SGGCN used SGNet-101 as backbone, which is built by ShuffleGhost Block (Huang et al., 2021) to extract features with a low computation cost. In order to make sufficient usage of the information in GCN, a new GCN architecture is designed to combine information from different layers together in GCNM module so that we can utilize various hierarchical features and meanwhile make the GCN scheme faster. The experiment on CheXPert datasets illustrated that SGGCN achieves a considerable performance. Compared with GCN and ResNet-101 (He et al., 2015) backbone (test AUC 0.8080, parameters 4.7M and FLOPs 16.0B), the SGGCN achieves 0.7831 (−3.08%) test AUC with parameters 1.2M (−73.73%) and FLOPs 3.1B (−80.82%), where GCN with MobileNet (Sandler and Howard, 2018) backbone achieves 0.7531 (−6.79%) test AUC with parameters 0.5M (−88.46%) and FLOPs 0.66B (−95.88%).

Neural network based fast prediction of βN limits in HL-2M

The artificial neural networks (NNs) are trained, based on the numerical database, to predict the no-wall and ideal-wall βN limits, due to onset of the n = 1 (n is the toroidal mode number) ideal external kink instability, for the HL-2M tokamak. The database is constructed by toroidal computations utilizing both the equilibrium code CHEASE and the stability code MARS-F. The stability results show that (i) the plasma elongation generally enhances both βN limits, for either positive or negative triangularity plasmas; (ii) the effect is more pronounced for positive triangularity plasmas; (iii) the computed no-wall βN limit linearly scales with the plasma internal inductance, with the proportionality coefficient ranging between 1 and 5 for HL-2M; (iv) the no-wall limit substantially decreases with increasing pressure peaking factor. Furthermore, both the Neural Network (NN) model and the Convolutional Neural Networks model (CNN) are trained and tested, resulting in consistent results. The trained NNs predict both the no-wall and ideal-wall limits with as high as 95% accuracy, compared to those directly computed by the stability code. Additional test cases, produced by the Tokamak Simulation Code (TSC), also show reasonable performance of the trained NNs, with the relative error being within 10%. The constructed database provides effective references for the future HL-2M operations. The trained NNs can be used as a real-time monitor for disruption prevention in the HL-2M experiments, or serve as part of the integrated modeling tools for ideal kink stability analysis.

Features Importance Analysis of Diesel Vehicles’ NOx and CO2 Emission Predictions in Real Road Driving Based on Gradient Boosting Regression Model

In order to have an accurate and fast prediction of the artificial intelligence (AI) model, the choice of input features is at least as important as the choice of model. The effect of input features selection on the emission models of light diesel vehicles driven on real roads was investigated in this paper. The gradient boosting regression (GBR) model was used to train and to predict the emissions of nitrogen oxide (NOx), carbon dioxide (CO2), and the fuel consumption of real driving diesel vehicles in urban scenarios, the suburbs, and on highways. A portable emissions measurement system (PEMS) system was used to collect data of vehicles as well as environmental conditions. The vehicle was run on two routes. The model was trained with the first route data and was used to predict the emissions of the second route. There were ten features related to the NOx model and nine features associated with the CO2 model. The importance of each feature was sorted, and a different number of features were used as input to train the models. The best NOx model had the coefficient of determination (R2) values of 0.99, 0.99, and 0.99 in each driving pattern (urban, suburbs, and highways). Predictions of the second route had the R2 values of 0.88, 0.89, and 0.96 respectively. The best CO2 model had the R2 values of 0.98, 0.99, and 0.99 in each driving pattern, respectively. Predictions of the second route had the R2 values are 0.79, 0.82, and 0.83, respectively. The most important features for the NOx model are mass air flow rate (g/s), exhaust flow rate (m3/min), and CO2 (ppm), while the important features for the CO2 model are exhaust flow rate (m3/min) and mass air flow rate (g/s). It is noted that the regression models based on the top three features may give predictions very close to the measured data.

Data-Driven Modeling of Geometry-Adaptive Steady Heat Convection Based on Convolutional Neural Networks

Heat convection is one of the main mechanisms of heat transfer, and it involves both heat conduction and heat transportation by fluid flow; as a result, it usually requires numerical simulation for solving heat convection problems. Although the derivation of governing equations is not difficult, the solution process can be complicated and usually requires numerical discretization and iteration of differential equations. In this paper, based on neural networks, we developed a data-driven model for an extremely fast prediction of steady-state heat convection of a hot object with an arbitrary complex geometry in a two-dimensional space. According to the governing equations, the steady-state heat convection is dominated by convection and thermal diffusion terms; thus the distribution of the physical fields would exhibit stronger correlations between adjacent points. Therefore, the proposed neural network model uses convolutional neural network (CNN) layers as the encoder and deconvolutional neural network (DCNN) layers as the decoder. Compared with a fully connected (FC) network model, the CNN-based model is good for capturing and reconstructing the spatial relationships of low-rank feature spaces, such as edge intersections, parallelism, and symmetry. Furthermore, we applied the signed distance function (SDF) as the network input for representing the problem geometry, which contains more information compared with a binary image. For displaying the strong learning and generalization ability of the proposed network model, the training dataset only contains hot objects with simple geometries: triangles, quadrilaterals, pentagons, hexagons, and dodecagons, while the testing cases use arbitrary and complex geometries. According to the study, the trained network model can accurately predict the velocity and temperature field of the problems with complex geometries, which has never been seen by the network model during the model training; and the prediction speed is two orders faster than the CFD. The ability of accurate and extremely fast prediction of the network model suggests the potential of applying reduced-order network models to the applications of real-time control and fast optimization in the future.

Fast Prediction for Resistance Spot Welding Deformation Using Inherent Strain Method and Nugget Model

In a car body, there exist thousands of resistance spot welds, which may induce large deformation during the manufacturing process. Therefore, it is expected that automotive industries will develop a method and a computing system for the fast and simple prediction of its deformation. Although the inherent strain method has been used for the fast prediction of arc welding deformation, it has not been applied to resistance spot welding so far. Additionally, the electrical-thermal-mechanical coupling analysis for the deformation induced by resistance spot welding is complicated and much more time-consuming. Therefore, in this study, a nugget model of the resistance spot weld has been developed, and the inherent strain method is extended for use in the fast prediction of resistance spot welding deformation. In addition, the deformation of a vehicle part with 23 resistance spot welds was efficiently predicted within around 90 min using the inherent strain method, displaying good accuracy compared with the measurement.