• The Journal of Korea Robotics Society
• /
• v.18 no.2
• /
• pp.143-154
• /
• 2023

In this study, a fast motion planning method for the swing motion of a 6x6 wheel-legged robot to traverse large obstacles and gaps is proposed. The motion planning method presented in the previous paper, which was based on trajectory optimization, took up to tens of seconds and was limited to two-dimensional, structured vertical obstacles and trenches. A deep neural network based on one-dimensional Convolutional Neural Network (CNN) is introduced to generate keyframes, which are then used to represent smooth reference commands for the six leg angles along the robot's path. The network is initially trained using the behavioral cloning method with a dataset gathered from previous simulation results of the trajectory optimization. Its performance is then improved through reinforcement learning, using a one-step REINFORCE algorithm. The trained model has increased the speed of motion planning by up to 820 times and improved the success rates of obstacle crossing under harsh conditions, such as low friction and high roughness.

• Geomechanics and Engineering
• /
• v.38 no.4
• /
• pp.439-453
• /
• 2024

The research findings of two nonlinear machine learning and soft computing models- the Cuckoo optimization algorithm (COA) and the Teaching-learning-based optimization (TLBO) in combination with artificial neural network (ANN)-are presented in this article. Detailed finite element modeling (FEM) of a shallow footing on two layers of cohesionless soil provided the data sets. The models are trained and tested using the FEM outputs. Additionally, various statistical indices are used to compare and evaluate the predicted and calculated models, and the most precise model is then introduced. The most precise model is recommended to estimate the solution after the model assessment process. When the anticipated findings are compared to the FEM data, there is an excellent agreement, which indicates that the TLBO-MLP solutions in this research are reliable (R²=0.9816 for training and 0.99366 for testing). Additionally, the optimized COA-MLP network with a swarm size of 500 was observed to have R² and RMSE values of (0.9613 and 0.11459) and (0.98017 and 0.09717) for both the normalized training and testing datasets, respectively. Moreover, a straightforward formula for the soft computing model is provided, and an excellent consensus is attained, indicating a high level of dependability for the suggested model.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.39 no.5
• /
• pp.433-441
• /
• 2011

The optimization of a centrifugal compressor was conducted. The ANN (Artificial Neural Network) was adopted as an optimization algorithm, and it was learned and trained with the DOE (Design of Experiment). In the DOE, it was predicted the main effect and the interaction effect of design variables to the objective function. The ANN was improved in the optimization process using the GA (Genetic Algorithm). When any output at each generation was reached a standard level, it was re-calculated by the CFD (Computational Fluid Dynamics) and it was applied to develop a new ANN. After 6th generation, the prediction difference between ANN and CFD was less than 1%. A pareto of the efficiency versus the pressure ratio was obtained through the 21th generation. Using this method, the computational time for the optimization was equivalent to the time consumed by the gradient method, and the optimized results of multi-objective function were obtained.

• International Journal of Naval Architecture and Ocean Engineering
• /
• v.7 no.4
• /
• pp.750-769
• /
• 2015

The geometry of engineering systems affects their performances. For this reason, the shape of engineering systems needs to be optimized in the initial design stage. However, engineering system design problems consist of multi-objective optimization and the performance analysis using commercial code or numerical analysis is generally time-consuming. To solve these problems, many engineers perform the optimization using the approximation model (response surface). The Response Surface Method (RSM) is generally used to predict the system performance in engineering research field, but RSM presents some prediction errors for highly nonlinear systems. The major objective of this research is to establish an optimal design method for multi-objective problems and confirm its applicability. The proposed process is composed of three parts: definition of geometry, generation of response surface, and optimization process. To reduce the time for performance analysis and minimize the prediction errors, the approximation model is generated using the Backpropagation Artificial Neural Network (BPANN) which is considered as Neuro-Response Surface Method (NRSM). The optimization is done for the generated response surface by non-dominated sorting genetic algorithm-II (NSGA-II). Through case studies of marine system and ship structure (substructure of floating offshore wind turbine considering hydrodynamics performances and bulk carrier bottom stiffened panels considering structure performance), we have confirmed the applicability of the proposed method for multi-objective side constraint optimization problems.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.2
• /
• pp.77-86
• /
• 2004

In general, small-sized neural networks, even though they show good generalization performance, tend to fail to team the training data within a given error bound, whereas large-sized ones learn the training data easily but yield poor generalization. Therefore, a way of achieving good generalization is to find the smallest network that can learn the data, called the optimal-sized neural network. This paper proposes a new scheme for network pruning with ‘impact factor’ which is defined as a multiplication of the variance of a neuron output and the square of its outgoing weight. Simulation results of function approximation problems show that the proposed method is effective in regression.

• Proceedings of the KIEE Conference
• /
• 2005.10b
• /
• pp.152-154
• /
• 2005

In this paper, we propose a new architecture of Fuzzy Polynomial Neural Networks(FPNN) by means of genetically optimized Fuzzy Polynomial Neuron(FPN) and discuss its comprehensive design methodology involving mechanisms of genetic optimization, especially Genetic Algorithms(GAs). The design of the network exploits the extended Group Method of Data Handling(GMDH) with some essential parameters of the network being provided by the designer and kept fixed throughout the overall development process. This restriction may hamper a possibility of producing an optimal architecture of the model. The proposed FPNN gives rise to a structurally optimized network and comes with a substantial level of flexibility in comparison to the one we encounter in conventional FPNNs. It is shown that the proposed genetic algorithms-based Fuzzy Polynomial Neural Networks is more useful and effective than the existing models for nonlinear process. We experimented with Medical Imaging System(MIS) dataset to evaluate the performance of the proposed model.

• Smart Structures and Systems
• /
• v.22 no.2
• /
• pp.175-183
• /
• 2018

Time-series data often contain one of the most valuable pieces of information in many fields including manufacturing. Because time-series data are relatively cheap to acquire, they (e.g., vibration signals) have become a crucial part of big data even in manufacturing shop floors. Recently, deep-learning models have shown state-of-art performance for analyzing big data because of their sophisticated structures and considerable computational power. Traditional models for a machinery-monitoring system have highly relied on features selected by human experts. In addition, the representational power of such models fails as the data distribution becomes complicated. On the other hand, deep-learning models automatically select highly abstracted features during the optimization process, and their representational power is better than that of traditional neural network models. However, the applicability of deep-learning models to the field of prognostics and health management (PHM) has not been well investigated yet. This study integrates the "residual fitting" mechanism inherently embedded in the wavelet transform into the convolutional neural network deep-learning structure. As a result, the architecture combines a signal smoother and classification procedures into a single model. Validation results from rotor vibration data demonstrate that our model outperforms all other off-the-shelf feature-based models.

• Nuclear Engineering and Technology
• /
• v.52 no.11
• /
• pp.2678-2685
• /
• 2020

The pipe bends and elbows in nuclear power plants (NPPs) are vulnerable to degradation mechanisms and can cause wall-thinning defects. As it is difficult to detect both the defects generated inside the wall-thinned pipes and the preliminary signs, the wall-thinning defects should be accurately estimated to maintain the integrity of NPPs. This paper proposes a deep fuzzy neural network (DFNN) method and estimates the collapse moment of wall-thinned pipe bends and elbows. The proposed model has a simplified structure in which the fuzzy neural network module is repeatedly connected, and it is optimized using the least squares method and genetic algorithm. Numerical data obtained through simulations on the pipe bends and elbows with extrados, intrados, and crown defects were applied to the DFNN model to estimate the collapse moment. The acquired databases were divided into training, optimization, and test datasets and used to train and verify the estimation model. Consequently, the relative root mean square (RMS) errors of the estimated collapse moment at all the defect locations were within 0.25% for the test data. Such a low RMS error indicates that the DFNN model is accurate in estimating the collapse moment for wall-thinned pipe bends and elbows.

• Bulletin of the Korean Chemical Society
• /
• v.28 no.9
• /
• pp.1472-1476
• /
• 2007

Artificial neural networks (ANNs) are successfully developed for the modeling and prediction of normalized polarity parameter (ETN) of 216 various solvents with diverse chemical structures using a quantitative-structure property relationship. ANN with architecture 5-9-1 is generated using five molecular descriptors appearing in the multi-parameter linear regression (MLR) model. The most positive charge of a hydrogen atom (q+), total charge in molecule (qt), molecular volume of solvent (Vm), dipole moment (μ) and polarizability term (πI) are input descriptors and its output is ETN. It is found that properly selected and trained neural network with 192 solvents could fairly represent the dependence of normalized polarity parameter on molecular descriptors. For evaluation of the predictive power of the generated ANN, an optimized network is applied for prediction of the ETN values of 24 solvents in the prediction set, which are not used in the optimization procedure. Correlation coefficient (R) and root mean square error (RMSE) of 0.903 and 0.0887 for prediction set by MLR model should be compared with the values of 0.985 and 0.0375 by ANN model. These improvements are due to the fact that the ETN of solvents shows non-linear correlations with the molecular descriptors.

• Journal of Microbiology and Biotechnology
• /
• v.22 no.7
• /
• pp.939-946
• /
• 2012

Antibiotic production with Streptomyces sindenensis MTCC 8122 was optimized under submerged fermentation conditions by artificial neural network (ANN) coupled with genetic algorithm (GA) and Nelder-Mead downhill simplex (NMDS). Feed forward back-propagation ANN was trained to establish the mathematical relationship among the medium components and length of incubation period for achieving maximum antibiotic yield. The optimization strategy involved growing the culture with varying concentrations of various medium components for different incubation periods. Under non-optimized condition, antibiotic production was found to be $95{\mu}g/ml$, which nearly doubled ($176{\mu}g/ml$) with the ANN-GA optimization. ANN-NMDS optimization was found to be more efficacious, and maximum antibiotic production ($197{\mu}g/ml$) was obtained by cultivating the cells with (g/l) fructose 2.7602, $MgSO_4$ 1.2369, $(NH_4)_2PO_4$ 0.2742, DL-threonine 3.069%, and soyabean meal 1.952%, for 9.8531 days of incubation, which was roughly 12% higher than the yield obtained by ANN coupled with GA under the same conditions.