Search

Cloud manufacturing is a consumer-centric requirement-driven manufacturing paradigm that integrates distributed resources for providing services to consumers in an on-demand manner. Scheduling of multiple tasks is an important technical means for satisfying consumer requirements in cloud manufacturing. However, high individualised requirements and the associated complex task structures complicate the task scheduling in cloud manufacturing. This paper establishes a more comprehensive model for scheduling multiple distinct tasks with complicated manufacturing processes. The hierarchical relationships (a mixture of dependency and independency) of subtasks within tasks are considered. The objectives involve three kinds of time and cost factors, namely processing time, setup time, transfer time and the respective cost. In addition, service quality is also considered into the optimisation objective. Two multi-objective-meta-heuristic algorithms, i.e. ACO-based multi-objective algorithm (MACO) and NSGA-II-based multi-objective algorithm (MGA), are designed to solve the scheduling problem. A detailed analysis of the performance of the two algorithms is performed by applying them to several different scheduling instances. Experimental results indicate that in most cases the MACO algorithm can obtain a more diverse set of Pareto solutions hence offering more alternatives to meet widely different users' needs. [ABSTRACT FROM AUTHOR]

With the advance of information technology, many fields have begun using data clustering to reveal data structures and obtain useful information. Most of the existing clustering algorithms are susceptible to outliers and noises as well as the initial solution. The fuzzy c-ordered-means (FCOM) method can handle outlier and noise problems by using Huber's M-estimators and Yager's OWA operator to enhance its robustness. However, the result of the FCOM algorithm is still unstable because its initial centroids are randomly generated. Besides, the attributes' weight also affect the clustering performance. Thus, this study first proposed an intuitionistic fuzzy weighted c-ordered-means (IFWCOM) algorithm that combines intuitionistic fuzzy sets (IFSs), the feature-weighted and FCOM together to improve the clustering result. Moreover, this study proposed a real-coded genetic algorithm-based IFWCOM (GA-IFWCOM) that employs the genetic algorithm to exploit the global optimal solution of the IFWCOM algorithm. Twelve benchmark datasets were used for verification in the experiment. According to the experimental results, the GA-IFWCOM algorithm achieved better clustering accuracy than the other clustering algorithms for most of the datasets. [ABSTRACT FROM AUTHOR]