Kumar Shivam

Dr. Kumar Shivam

Kun Shan University, Taiwan
+1
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 April 2020 in Energies
Verified
Journal Article
Open access
PDF file
KSKumar Shivam
Kumar Shivam
JTJong Chyuan Tzou
Jong Chyuan Tzou
SWShang-Chen Wu
Shang-Chen Wu
Wind energy is the most used renewable energy worldwide second only to hydropower. However, the stochastic nature of wind speed makes it harder for wind farms to manage the future power production and maintenance schedules efficiently. Many wind speed prediction models exist that focus on advance neural networks and/or preprocessing techniques to improve the accuracy. Since most of these models require a large amount of historic wind data and are validated using the data split method, the application to real-world scenarios cannot be determined. In this paper, we present a multi-step univariate prediction model for wind speed data inspired by the residual U-net architecture of the convolutional neural network (CNN). We propose a residual dilated causal convolutional neural network (Res-DCCNN) with nonlinear attention for multi-step-ahead wind speed forecasting. Our model can outperform long-term short-term memory networks (LSTM), gated recurrent units (GRU), and Res-DCCNN using sliding window validation techniques for 50-step-ahead wind speed prediction. We tested the performance of the proposed model on six real-world wind speed datasets with different probability distributions to confirm its effectiveness, and using several error metrics, we demonstrated that our proposed model was robust, precise, and applicable to real-world cases.

25 March 2020 in Applied Sciences
Verified
Journal Article
Open access
PDF file
JHJen-Chung Hsiao
Jen-Chung Hsiao
KSKumar Shivam
Kumar Shivam
CCC. L. Chou
C. L. Chou
Tai KamTai Kam
Tai Kam
In the design optimization of robot arms, the use of simulation technologies for modeling and optimizing the objective functions is still challenging. The difficulty is not only associated with the large computational cost of high-fidelity structural simulations but also linked to the reasonable compromise between the multiple conflicting objectives of robot arms. In this paper we propose a surrogate-based evolutionary optimization (SBEO) method via a global optimization approach, which incorporates the response surface method (RSM) and multi-objective evolutionary algorithm by decomposition (the differential evolution (DE ) variant) (MOEA/D-DE) to tackle the shape design optimization problem of robot arms for achieving high speed performance. The computer-aided engineering (CAE) tools such as CAE solvers, computer-aided design (CAD) Inventor, and finite element method (FEM) ANSYS are first used to produce the design and assess the performance of the robot arm. The surrogate model constructed on the basis of Box–Behnken design is then used in the MOEA/D-DE, which includes the process of selection, recombination, and mutation, to optimize the robot arm. The performance of the optimized robot arm is compared with the baseline one to validate the correctness and effectiveness of the proposed method. The results obtained for the adopted example show that the proposed method can not only significantly improve the robot arm performance and save computational cost but may also be deployed to solve other complex design optimization problems.

15 May 2020 in Energies
Verified
Journal Article
Open access
PDF file
KSKumar Shivam
Kumar Shivam
JTJong Chyuan Tzou
Jong Chyuan Tzou
SWShang-Chen Wu
Shang-Chen Wu
Increased concerns over global warming and air pollution has pushed governments to consider renewable energy as an alternative to meet the required energy demands of countries. Many government policies are deployed in Taiwan to promote solar and wind energy to cope with air pollution and self-dependency for energy generation. However, the residential sector contribution is not significant despite higher feed-in tariff rates set by government. This study analyzes wind and solar power availability of four different locations of southern Taiwan, based on the Köppen–Geiger climate classification system. The solar–wind hybrid system (SWHS) considered in this study consists of multi-crystalline photovoltaic (PV) modules, vertical wind turbines, inverters and batteries. Global reanalysis weather data and a climate-based electricity load profile at a 1-h resolution was used for the simulation. A general framework for multi-objective optimization using this simulation technique is proposed for solar–wind hybrid system, considering the feed-in tariff regulations, environmental regulations and installation area constraints of Taiwan. The hourly load profile is selected using a climate classification system. A decomposition-based differential evolutionary algorithm is used for finding the optimal Pareto set of two economic objectives and one environmental objective with maximum installation area and maximum PV capacity constraints. Two types of buildings are chosen for analysis at four climate locations. Analysis of Pareto sets revealed that the photovoltaic modules are economic options for a grid-connected mode at all four locations, whereas solar–wind hybrid systems are more environmentally friendly. A method of finding the fitness index for the Pareto front sets and a balanced strategy for choosing the optimal configuration is proposed. The proposed balanced strategy provides savings to users—up to 49% for urban residential buildings and up to 32% for rural residential buildings with respect to buildings without a hybrid energy system (HES)—while keeping carbon dioxide (CO2) emissions lower than 50% for the total project lifecycle time of 20 years. The case study reveals that for all four locations and two building types an HES system comprising a 15 kW photovoltaic system and a small capacity battery bank provides the optimal balance between economic and environmental objectives.

17 December 2020 in IEEE Access
Verified
Journal Article
Open access
JHJen-Chung Hsiao
Jen-Chung Hsiao
KSKumar Shivam
Kumar Shivam
ILI-Fang Lu
I-Fang Lu
TKTai-Yan Kam
Tai-Yan Kam
Robot calibration of an industrial robot is of importance for those applications requiring high positioning accuracy. The use of numerical optimization techniques to identify the accurate kinematic parameters and tackle the residual errors for improving positioning accuracy is still challenging. The difficulties in using the conventional optimization techniques is associated with the expensive computation of the second derivative terms in the Hessian matrix and even more linked to the additional improvement of the residual errors induced by the configuration and payload effects for the practical applications. A hybrid calibration procedure, which is robust and efficient, is proposed to handle the said problems in this study. The first step is to utilize the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm together with the quasi-Newton method to optimize the cost function with non-smooth nature established using massive measurement data. In the second step, two kinds of artificial neural network (ANN) algorithms are employed to further improve the nonlinear residual errors induced by payload and configuration effects. After the completion of the hybrid calibration, improvements of 88.1% and 80.1% have been attained, respectively, for the mean and maximum positioning errors of the robot end-effector, i.e., the mean/maximum positioning errors are reduced from 2.613mm/6.294mm to 0.310mm/1.255mm for 840 untrained measurement data. The experimental results also validated the effectiveness of the proposed method.

1 April 2020 in Journal of Physics: Conference Series
Verified
Conference Paper
Open access
JHJen-Chung Hsiao
Jen-Chung Hsiao
KSKumar Shivam
Kumar Shivam
TKT Y Kam
T Y Kam
Worm gearboxes are popular across various industrial applications since they offer significant gear ratios in small installation spaces. Despite having multiple advantages, worm gearboxes are subjected to higher friction due to sliding design and are prone to damage and increased transmission error over the period of operation. Delayed diagnosis of worm gearbox degradation can lead to low-quality products and/or unnecessary production line downtime. Using vibration characteristics of worm gearbox, it is possible to determine the fault and transmission error at a given period in time. In this paper, an ensemble machine learning model is trained and deployed to monitor the transmission error of worm gearbox and classify between new, operational and old conditions. 1D CNN (one dimensional convolutional neural network) model is used to automatically extract features in vibration signal of X, Y, and Z axes and predict the relevant state of worm gear. The proposed technique uses ensemble machine learning technique fusion of features extracted by multi-layer 1D CNN for three axes vibration data. The proposed method could achieve 96% accuracy and performs significantly better than traditional sequential and ensemble machine learning models on a dataset of with 7,870 samples with 800 samples labeled as new condition, 4,740 samples as operational and 2,330 as old.

7 April 2023
Verified
Conference Paper
Closed access
KSKumar Shivam
Kumar Shivam
SMSaikat Majumder
Saikat Majumder
Sumit ChakravartySumit Chakravarty
Sumit Chakravarty
Fingerprint-based indoor localization methods rely on a database of Received Signal Strength (RSS) measurements and corresponding location labels. However, collecting and maintaining such a database can be costly and time consuming. In this work, we proposed Wasserstein Generative Adversarial Networks (WGAN) to generate synthetic data for fingerprinting-based indoor localization. The proposed system consists of a WGAN that is trained on a dataset of real RSS measurements and corresponding location labels. The generator of the WGAN learns to generate synthetic RSS measurements, and the critic learns to differentiate the generated and the real measurements. We validate the proposed system on a dataset of real RSS measurements. The result of the proposed system shows better localization accuracy as compared to using real data, while being more cost-effective and scalable.