Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

 
Session Overview
Session
SES 6.2: Digital Product and Process Development
Time:
Wednesday, 28/Jun/2017:
1:50pm - 3:10pm

Session Chair: Mika Lohtander
Location: Aula N (first floor)

Show help for 'Increase or decrease the abstract text size'
Presentations

120. A Cloud-based Kanban Decision Support System for Resource Scheduling & Management

Krishnan Krishnaiyer, F. Frank Chen

The University of Texas at San Antonio, United States of America

For several decades, lean manufacturing methodologies have been used for manufacturing enterprise improvement particularly in operations and supply chain management. As these improvements evolve so does the complexity and the size of data. With the ubiquity of data and the scale of machine automation, abilities for rapid decision making and handling of ever increasing complexity of systems become necessary. The purpose of this research is to demonstrate how a cloud-based Kanban decision support system combined with a robust continuous improvement methodology can help operation managers to make an efficacious decision. Various applications in the literature infer Kanban as a method to control inventory. In this paper, we propose a novel method, Estimated, Actual and Total (EAT) Kanban Decision Support System (DSS) that can be used in any dashboard type monitoring of processes. The paper addresses two research questions: (1) How a robust cloud kanban decision support system will work for a service industry, particularly in resource scheduling and management? (2) Can a proof of concept implementation be scalable across operations management in various sectors? We share successful prototype implementation in Direct Mail Marketing, and Educational Testing product scheduling. The results indicated a dramatic reduction in scheduling time (from 180 minutes to 3 minutes) and the number of tools used (Consolidated 157 spreadsheets into 1 database).


11. Hybrid simulation for complex manufacturing value-chain environments

Cátia Sofia Rodrigues Barbosa1,2, Americo Azevedo1,2

1Inesc Tec, Portugal; 2Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, s/n 4200-465 Porto, Portugal

Simulation is very popular for modelling complex systems. Recent demands from global business optimization, integration of human decision making, and increased complexity of modern systems, push researchers for combining different simulation methods and getting deeper understanding into complex interactions between processes of very different nature, calling for hybrid simulation approaches. These approaches combine at least two of three simulation methods – System Dynamics, Discrete Event Simulation, and Agent Based Simulation.

Even though there is a growing interest in hybrid simulation, many questions remain unsolved, as the lack of a unified use of terms and definitions in the literature, which introduces ambiguity. Literature in hybrid simulation is very sparse, hampering the work of researchers interested in the topic. Many challenges concern authors when using more than one simulation method, as establishing information sharing between the models, converting time units for proper information exchange, the skills required for building the models, among others.

This work aims at providing insight on the use of hybrid simulation in the context of business, supply chains (SCs), manufacturing, and logistics; and the most important advantages and challenges of using hybrid simulation. We try to answer two research questions:

RQ1: How has hybrid simulation been used in business, SCs, manufacturing, and logistics?

RQ2: Which are the challenges and benefits of hybrid simulation compared to standalone simulation?

This paper reviews literature related to hybrid simulation, focusing on different combinations of methods and the advantages and challenges of using hybrid simulations.

For the structured literature review, a set of keywords considering the dispersion of terms in the literature was selected, and three databases (Scopus, Science Direct, and Emerald Insight) were chosen. The papers were filtered based on abstract and full-text reading. Forward and backward search were used for increasing the range of papers analysed. More than 50 papers were fully analysed, across more than 20 years of publications.

Despite all the papers fully analysed targeted hybrid simulations, the modes of operation and relationships established between the models differ. Therefore, it is relevant understanding the different approaches to hybrid simulations. We present a classification scheme for the analysed papers, based on the classification scheme proposed by Swinerd and McNaught (2012), which included interfaced, sequential and integrated classifications, and adding the enrichment taxonomy as in Morgan, Howick and Belton (2016).

Even though hybrid simulation approaches are more frequent, combining two methods is only justified when the developed models are of equal importance to the overall goal of the simulation. Combining models using different methods requires effort and precision to establish information sharing. Common problems which arise include the different time units used in the models. Hybrid models require knowledge about different simulation methods; high skills and flexibility. In spite of the high demands of hybrid simulation, many advantages can be achieved. One of the benefits of hybrid simulation is flexibility. It is possible to simulate different levels of aggregation, avoiding problems of model consistency. Among others, hybridism allows using complementary methods, coupling methods, and exploration of multilateral problems.


90. New Approaches for the Determination of Specific Values for Process Models in Machining Using Artificial Neural Networks

Frank Arnold, Albrecht Hähnel, Andreas Nestler, Alexander Brosius

TU Dresden, Germany

The acceptance of the use of mathematical models for the determination of processforces is directly dependent on the quality of the characteristic values used. Especially in machining, the quality depends on the available information on the entire system machine-tool-material. The permanent development on the system components as well as the use of innovative processing strategies and new methods for processing simulation are drivers of development. An application of powerful mathematical models only makes sense if the specific characteristic values necessary for the process model are present and also up-to-date. In order to ensure this up-to-date, considerable effort is required to determine these variables. This time and cost implementation makes the application of process models unattractive in industrial applications. For the determination of the specific cutting forces of the cutting force model according to Kienzle, machining tests have to be prepared, carried out and evaluated. This requires the knowledge of an expert as well as the use of additional measuring technology. As a rule, these expenditures are not operated, which means that the available potentials in the overall system machine-tool-material are not used extensively. The approach of automated data acquisition without the need for additional measuring technology in the cutting machine is one possibility of a broader application. Modern CNC offer extensive information and communication functions. A concept for the detection of selected dynamic process data is developed and implemented using the example of the determination of specific cutting forces for the cutting force model according to Kienzle on the application of 2.5D milling. The processing of the discrete process data, which is recorded directly at the CNC, with different mathematical approaches is investigated and evaluated. Special attention is given to a separation of the signals into the components from the basic behavior of the machine and the fractions from the machining process itself as well as a clear detection for an automated evaluation. The following illustration of the recorded process data on the physical variables allows the determination of the specific cutting forces. This process can be carried out concurrently and is not a significant additional effort. Using the machine learning with artificial neural networks using the ability to generalize, specific characteristic values are determined on the basis of process data. This allows mathematical models to be supplied with current characteristic values over a wide range of applications. A major application is seen in the improved design of machining processes.


229. The Evaluation of Resonance Frequency for Piezoelectric Transducers by Machine Learning Methods

Fengming Chang

National Taitung Jr College, Taiwan

A piezoelectric transducer is a component employed in the applications of transmitting and receiving of sound wave. The distance of the sound wave could send is determined by the transducer frequency. Therefore, to measure the frequency becomes an important issue. However, it needs a lot of experiments to simulate and measure the transducer’s resonance frequency in the laboratory. To solve this problem, this research estimate a transducer’s frequency by machine learning methods instead of a laboratory experiment. The proposed method are compared with other methods, such as artificial neural network, support vector machine, C4.5, neuro-fuzzy, and mega-fuzzification. The results show that machine learning methods are efficiency ways to assess the resonance frequency of a piezoelectric transducer. Besides, mega-fuzzification method has the best accuracy among the comparative methods in this case.



 
Contact and Legal Notice · Contact Address:
Conference: FAIM 2017
Conference Software - ConfTool Pro 2.6.110
© 2001 - 2017 by H. Weinreich, Hamburg, Germany