Approaches for predicting financial markets, including conventional statistical methods and recent deep learning methods, have been investigated in many studies. However, financial time series data (e.g., daily stock market index) contain noises that prevent stable predictive model learning. Using these noised data in predictions results in performance deterioration and time lag. This study proposes paddingbased Fourier transform denoising (P-FTD) that eliminates the noise waveform in the frequency domain of financial time series data and solves the problem of data divergence at both ends when restoring to the original time series. Experiments were conducted to predict the closing prices of S&P500, SSE, and KOSPI by applying data, from which noise was removed by P-FTD, to different deep learning models based on time series. Results show that the combination of the deep learning models and the proposed denoising technique not only outperforms the basic models in terms of predictive performance but also mitigates the time lag problem.

Over the past decades, additive manufacturing has rapidly advanced due to its advantages in enabling diverse material usage and complex design production. Nevertheless, the technology has limitations in terms of quality, as printed products are sometimes different from their desired designs or are inconsistent due to defects. Warping deformation, a defect involving layer shrinkage induced by the thermal residual stress generated during manufacturing processes, is a major factor in lowering the quality and raising the cost of printed products. This study utilized a variety of thermal time series data and the K-nearest neighbors (KNN) algorithm with dynamic time warping (DTW) to detect and predict the warping deformation in the printed parts using fused deposition modeling (FDM) printers. Multivariate thermal time series data extracted from thermocouples were trained using DTW-based KNN to classify warping deformation. The results showed that the proposed approach can predict warping deformation with an accuracy of over 80% by only using thermal time series data corresponding to 20% of the whole printing process. Additionally, the classification accuracy exhibited the promising potential of the proposed approach in warping prediction and in actual manufacturing processes, so the additional time and cost resulting from defective processes can be reduced.

The development of lattice design approaches using topology optimization to create complex lattice structures is the focus of ongoing research as the lattice structures show great potential throughout a wider range of research fields. Unfortunately, many of the methods suggested are often very difficult to industrialize due to their complexity and excessive computational costs in the design process. This paper proposes a novel framework of generating non-periodic lattice structures using topologically pre-optimized building blocks to improve the computational efficiency of the optimization process while maintaining the performance of the target structure and to make the process of obtaining optimal design more accessible for people with limited knowledge of topology optimization. The proposed method is comprised of two phases: the generation of a library holding optimized building blocks that can endure various loading conditions and the allocation of these building blocks at the proper location in the design space based on stress information obtained from the finite element analysis. Numerical examples are presented to show that the method is feasible and able to achieve reasonable performances, demonstrating its efficacy.

No abstract

No abstract