Mohsen Amra

Master's in Industrial Engineering


Publications Abstract

A. Journal

A.1. Published Articles

Flexible job-shop scheduling problem considering machine and order acceptance, transportation costs, and setup times

This paper, for the first time, studied a new extension of the flexible job shop scheduling problem by assuming the acceptance and rejection of machines and orders. The flexible job shop problem was extended to implement production without a factory in natural environments. Therefore, the mixed-integer linear programming (MILP) model was developed for this problem aiming to minimize total costs, including the fixed cost of machine selection, variable operational cost, transportation cost, and order rejection cost. Due to the high complexity of this problem, a heuristic algorithm was employed to find an acceptable solution. For algorithm performance evaluation, 40 samples were randomly generated and solved using the mathematical model and the proposed algorithm. The results of analyzing random samples showed a negligible error rate indicating algorithm efficiency.

Ziaee, M., Mortazavi, J. & Amra, M. Flexible job shop scheduling problem considering machine and order acceptance, transportation costs, and setup times. Soft Comput (2021).

A.2. Under Review Articles

  • Formation of Optimal Multi-Period Mean-Semivariance Portfolio with Fuzzy Interest Rates: Case Study on the New York Stock Exchange.

This study aimed to find a mechanism for allocating assets between a desired collection of stocks with maximum return and minimum risk at the same time. It was also tried to use the uncertain interest rate for each stock to optimize the portfolio. To this end, a bi-objective multi-period mean-semivariance was first used to model the major problem. Then, the model was investigated under the primary state by considering the existing constraints. This study investigated the effect of an uncertain and fuzzy interest rate on portfolio composition in the New York Stock Exchange. The results showed that the multi-period mean-semivariance model with a fuzzy interest rate can be used for portfolio optimization. Portfolios composed by this model are more efficient. Due to uncertain interest rate, such portfolios are more optimal than the portfolios composed by other methods and algorithms.

  • Lagrange relaxation with Fix-and-Optimize strategy for a robust, resilient and sustainable closed-loop supply chain network under risk-aversion.

Nowadays, the research on the resiliency and sustainability dimensions of closed-loopsupply chains, is very active topic. The present study considers a resilient andsustainable closed-loop supply chain network design with the use of robustoptimization and risk aversion (RRSCLSCNDR). A two-stage stochastic programmingmodel is developed to formulate the proposed problem. The objective functions are theminimization of total costs, CO2 emissions, and energy consumption whilemaximization of employment as the social objective. Another contribution is the use ofrobust optimization with Conditional Value at Risk (CVaR) criterion for risk reduction toachieve reliability. The MINIMAX method with Entropic Value at Risk (EVaR) strategyis used to handle the proposed model. An Lp-Metric method is used to merge theobjectives to transform the multi-objective model into a single objective one. Since thecomplexity of this model is high, the Lagrange relaxation algorithm with a heuristicstrategy called Fix-and-Optimize rule, is applied to find a feasible lower bound and anoptimal upper bound in large-scale tests. To show the applicability of this research, acase study of a car assembler company in Iran is considered. The results confirm thatthe proposed model provides a better estimation of costs, environmental pollution,energy consumption, and employment level compared with the state of the art studiesand the proposed heuristic encourages further analyses and the development of thisalgorithm for other closed-loop supply chain problems.

B. Conference

A.1. Published Articles

A combined deep CNN-LSTM Network for classifying T1 weighted Magnetic Resonance Brain tumor

One of the main challenges in treating tumors and assessing disease progression is diagnosing tumor size And distinguish tumor types from each other. Manual tumor segmentation in three-dimensional Magnetic Resonance images (volume MRI) is a time-consuming and tedious task. Its accuracy depends heavily on the operator’s experience doing it. The need for an accurate and fully automatic method for segmenting brain tumors and measuring tumor size is strongly felt. This paper first uses a combined CNN-LSTM method to detect HG and LG tumors in ۳D brain images. Then it used the UNET Neural Network to improve the location of the tumor in the brain. In this article, we use BRATS ۲۰۱۸ database images, And manual segmentation is used as the Grand truth. in this paper, we showed that the proposed method could effectively perform segmentation.

Khoshkhooy Titkanlou, Maryam and Kazemi, Mostafa and Amra, Mohsen,1400,A combined deep Neural Network for classifying T۱ weighted Magnetic Resonance Brain tumor,Twelfth International Conference on Information Technology, Computer and Telecommunications,