Doğuş Üniversitesi Dergisi, 4 (2) 2003, 207-216 MODELING AUTOMATED GUIDED VEHICLE SYSTEMS IN
MATERIAL HANDLING
OTOMATİKLEŞTİRİLMİŞ REHBERLİ ARAÇ SİSTEMLERİNİN
TRANSPORT TEKNİĞİNDE MODELLEMESİ
Bülent SEZEN
Gebze Yükk Teknoloji Enstitüsü, İşletme Fakültesi
ABSTRACT: The study objectives are to 1) provide information regarding the u and benefits of Automated Guided Vehicle (AGV) systems in manufacturing environments, and 2) review the literature related to design, modeling and simulation of AGV systems. We classify the tools utilized in design problems of AGV systems as analytical and simulation-bad tools. Then, give examples of both categories from related literature.
Keywords: Automated Guided Vehicle Systems Design, Modeling and Simulation.
ÖZET: Çalışmanın amaçları; 1) Otomatikleştirilmiş Rehberli Araç (ORA, ingilizcesi, Automated Guided Vehicle, AGV) sistemlerinin kullanımı ve faydaları hakkında bilgiler vermek ve 2) ORA sistemlerinin tasarım, modellenme ve simulasyonu (benzetimi) ile ilgili kapsamlı bir literatür incelemesinin sonuçlarını sunmaktır. Öncelikle ORA sistemlerinin tasarım problemlerinde kullanılan yöntemleri analitik ve simülasyon yöntemler olarak ikiye ayrılıp, daha sonra, ilgili literatürden her iki gruba ait örnekler verilmektedir.
Anahtar Kelimeler: Otomatik Rehberli Araç Sistemleri Tasarımı, Modelleme ve
Simülasyon.
1. Introduction
Material handling in manufacturing systems is becoming easier as the automated machine technology is improved. Today’s rapid developments in technology prents manufacturing firms a variety of alternatives for in-plant transportation. An Automated Guided Vehicle (AGV) system is such an advanced material handling system that involves a fleet of driverless vehicles (AGVs) which follow a guided path and are controlled by a computer (Hammond, 1986). The aim of this study is to 1) provide information regarding the u of AGV systems in manufacturing environments, and 2) revi
勤能补拙的例子
ew the literature related to design, modeling and simulation of AGV systems.
2. Automated Guided Vehicle (AGV) Systems
A typical AGV consists of the frame, batteries, electrical system, drive unit, steering, precision stop unit, on-board controller, communication unit, safety system, and work platform. AGV systems are mainly ud for distribution of materials in warehou environments, and movement of material to and from production areas and storage areas in manufacturing facilities. The first Automated Guided Vehicle
208
Bülent SEZEN
高速公路收费
(AGV) application was for transporting groceries in a warehou (Hammond, 1986). According to statistics in 1989 (Gould, 1990), AGV system installations with respect to their application types were profiled as following: JIT delivery systems (56%), FMS/FAS transfer system (13%), storage load transfer, non-AS/RS (12%), AS/RS interface (8%), progressive asmbly (7%), mini-load AS/RS interface (1%), and others (3%). Some other applications of AGV systems in non-manufacturing env
ironments include, but are not limited to, delivering mail, messages, and packages in offices, and delivering meals and laundry in hospitals.
AGVs can be ud in two different ways (Hammond, 1986). The first approach is to attach a workpiece to the AGV having all manufacturing process done while the AGV carries the workpiece from station to station. In this approach, the AGV is freed only after all the process for the workpiece are completed. The cond approach is to u vehicles only for moving the workpieces from one station to another. Vehicle is assigned to the workpiece only for a single trip. In the former, number of vehicles required is significantly greater than in a normal AGV system. General Motors Company is the pioneer of such an asmbly system built in the U.S. with 185 unit-load carriers.
AGV guidance techniques include wire guide path, optical guide path, and off-wire guidance. In the wire guide path technique, wires with varying frequencies are buried in the floor. AGVs lect a path at a control point according to the assigned frequency. In the optical guide path technique, an AGV focus a beam of light on a reflective tape or a painted strip and follows the path by measuring the amplitude of the reflected light. The disadvantages of wire guidepaths and optical guidepaths have caud the development of off-wire guidance techniques such as lar triangulation, floor-grid referencing, and gyroscopic guidance. The advantages of the techniques are that there is no nee
蒙被综合症d for floor cutting or painting, and the guidepaths are easily modified.
2.1. Basic Vehicle Types
Types of AGVs can be categorized as towing vehicles, pallet trucks, and unit-load carriers. Towing vehicles pull a ries of trailers that are attached to the vehicle. The trailers are attached to and detached from the vehicle manually at the stations. The vehicle does not have lifting capabilities nor a transfer mechanism. It can be ud for any type of load. Pallet trucks are ud for palletized loads and can have high lifting capabilities. They can pick up and deposit loads at the floor level. Unit-load carrier may carry single or multiple loads on their deck. Some are capable of traveling sideways. The transfer mechanism of the carrier can be either an active or passive conveyor, such as a roller, belt, or chain conveyor, or it may be a lift/lower deck.
2.2. Benefits of an AGV System
According to ca studies of AGV applications provided by the Material Handling Institute (1993), benefits of building and using AGV systems include labor costs saving, better schedule of WIP, flexible material handling, effective inventory
Modeling Automated Guided Vehicle Systems In Material Handling 209 control, greater quality assurance and safety, incread production, improved utilization of space, and flexible routing.
3. AGV Systems Design Problem
Typical objectives in design of AGV systems include 1) evaluation of the feasibility of an AGV system, 2) evaluation of the dispatching rules, 3) elimination of traffic problems, 4) maximizing the throughput, 5) maximizing the vehicle utilization, 6) minimizing the inventory level, 6) minimizing the transportation costs, and 7) maximizing the space utilization. Tools ud in AGV system design can be classified in two main categories: analytical tools and simulation-bad tools. Analytical tools are mathematical techniques such as queuing theory, integer programming, heuristic algorithm, and Markov Chains. A number of analytical approaches to the design of AGV systems have been propod in the literature.
3.1. Analytical tools
Tanchoco et al. (1987) compared the effectiveness of CAN-Q, an analytical model bad on queuing theory and ud for analyzing work flows through a manufacturing system, with a simulation-bad model built in AGVSim (Egbelu and Tanchoco, 1982). CAN-Q underestimated the actual number of
vehicles required. Their analysis indicates that the results obtained through CAN-Q provide a good starting point for the simulation study.
Bozer and Srinivasan (1991) introduced the concept of 'tandem configuration' to the design of AGV systems. The tandem configuration is bad on partitioning all of the workcenters into non-overlapping, single vehicle clod loops. It offers less complicated control systems, but has less tolerance for vehicle breakdowns and requires additional floor space. The authors also developed an analytical model to estimate the throughput capacity of a single vehicle in a clod loop. Mahadevan and Narendran (1993) developed an analytical model for estimation of the number of AGVs. They suggested to start with rough-cut analytical methods, followed by the u of sophisticated mathematical models and then to apply simulation if the level of complexity of the AGV system was high. As the number of parts in the system increas, the problem becomes intractable and needs to be analyzed by simulation method.
Johnson and Brandeau (1993) modeled an AGV system as a queuing system and the design model was formulated as a binary integer program with non-linear waiting time constraints. They then developed two different enumeration algorithms to solve the analytical model. Analytical models are generally bad on steady-state flow systems (Tanchoco, 1994). Therefore, analytical techniques m
ay fail when they are applied to real industrial cas. The techniques may give inaccurate estimates under random environments. In conclusion, analytical techniques may be best suited for obtaining initial estimates in the design of an AGV system (Egbelu, 1987).
3.2. AGV Systems Simulation
Simulation software that can be ud for AGV system simulation can be grouped in three categories (Tanchoco, 1994 : 1) General-purpo simulation languages (e.g.
210萨维尔街
Bülent SEZEN
SLAM II, SIMAN IV, etc.), 2) Simulation packages designed for the general simulation of manufacturing systems (e.g. SIMPLE++, AutoMod II, ProModel, SIMFACTORY II.5, etc.), and 3) Simulation software specially created for analyzing AGV systems by using general programming languages such as C, FORTRAN, BASIC, LISP, etc.
3.2.1. General-purpo Simulation Languages
家风家训有哪些
Several AGV system simulation models have been developed using general-purpo simulation languages such as SLAM II (Pritsker, 1995), SIMAN (Pegden et al., 1990), and GPSS/H (Henrikn and Crane, 1989). Seifert et al. (1995) developed a discrete-event simulation model written in SLAM II to analyze the operation of an AGV system under a variety of vehicle routing strategies. Their model handled multiple layouts and pedestrians in the system. It was a mixed-language model that was written in SLAM II with event-processing functions written in the C programming language. A specific performance measure was employed by their simulation model. It was the difference between AGV's actual travel time and the corresponding theoretical travel time of the AGV with respect to its speed and the travel length.
Ulgen and Kedia (1990) built a simulation model using SLAMSYSTEM to analyze the main effects of the design and operational variables on the performance of a cellular asmbly system employing AGVs. The factors and alternatives considered in their model were alternate track layout designs, the effect of scheduling rules, and the effect of different cycle time ratios. The measure of the system performance was the average throughput per shift. They stated that the simulation was easy to implement, especially, the scheduling rules in the SLAMSYSTEM.
Takakuwa (1993) created a simulation model in SIMAN to measure the cost effectiveness of large sc
ale AS/RS-AGV systems bad on the number of AGVs to install. First, the overall layout of the system was determined, and then, specifications of the system such as number of AGVs, number of conveyors, the buffer size on each conveyor, and so on, were defined. Their main system measure was the total flow time. Lee (1996) developed a discrete-event simulation model in SIMAN to evaluate the performance of a number of composite AGV dispatching rules which could be implemented in manufacturing or asmbly systems. The number of AGVs needed was determined bad on a preliminary simulation study. The performance measures were the throughput, flow time, and the WIP level. Although material handling features have been added to the general-purpo simulation languages, such as SLAM and SIMAN, the features do not provide sufficient flexibility to simulate the great diversity of many different material handling systems (King and Kim, 1995). Simulation of AGV systems can be more easily accomplished by starting with simulation packages specifically developed for manufacturing systems.
Modeling Automated Guided Vehicle Systems In Material Handling 211 3.2.2. Simulation Packages Specific for Manufacturing Systems
The cond category of simulation software includes some general-purpo manufacturing simulation systems such as AutoMod II, ProModel, XCELL+, SIMFACTORY II.5, SIMPLE++, etc.
Prasad and Rangaswami (1988) developed a graphic simulation model of an integrated miconductor sort, asmbly and test facility by using AutoMod as the primary tool. AutoMod is a 'macro' language which is bad on the GPSS simulation language. Two different AGV control systems, a global control system versus a local control system, have been analyzed with the simulation model. They stated that AGV bottlenecks, congestion and deadlocks could be easily identified by using AutoMod's animation feature.
Quinn (1985) created a simulation-bad system with AutoMod for development and testing of AGV control software. Data from a CAD system was ud to describe the guide path. Output from the model was interfaced to an emulator that imitated network protocol to controller for testing the software. A generic blocking scheme was ud in the model. Quinn (1985) stresd the fact that AGV vendors had unique blocking designs of their own.
Jayaraman (1993) developed an AGV system design for a company manufacturing antilock breaking systems by using ProModel. He compared manual transportation with AGV transportation. Input data to the system included a file containing the plant layout in AUTOCAD drawings of each asmbly cell, forecast requirements, processing times, material handling information, and the bills of materials. The system throughput and the AGV utilization were ud as the measure of system p
erformance. Dewsnup and Bollenbach (1995) discusd the u of ProModel for Windows for modeling AGV systems. They studied two parate but overlapping systems. The model needed special interction logic to avoid collisions. The objective was to determine the number of AGVs and to identify control logic for AGV system.
3.2.3. Simulation Software Specifically Created for AGV Simulation
There are some simulation software packages that are specifically developed for analyzing AGV systems. In this ction, we will briefly review the AGV simulation programs. The initial part of the ction will describe codes developed by traditional programming methodologies while the later part of the ction identifies efforts that are bad on object-oriented approaches.
3.2.3.1. Simulation Bad on Traditional Programming Approaches
Egbelu and Tanchoco (1982) developed AGVSim, a simulation package for designing AGV systems. The package was developed in the FORTRAN language. It was a tool for analyzing, planning, and designing AGV systems. In this package, a network was modeled as a collection of nodes and arcs. Only the unidirectional flow pattern was considered.
212
索尼投影仪Bülent SEZEN
Sinriech and Tanchoco (1992) ud AGVSim to evaluate the performance of a single loop guide path under different dispatching rules. They stresd the impact of empty vehicle flow on the system performance. AGVSim consists of two parate routines. The first routine calculates the shortest path between pairs of points in the network. The cond routine is the main simulator which executes the simulation and reports the results. AGVSim provides support programs to enter and remove data. Anderson (1985) created SattControl, an AGV simulation package, which was ud as a tool for planning and testing of an AGV route layout. Input parameters to the package included AGV track layout, number of pallets per hour, number of AGVs, loading/unloading times, AGV speed (parate for loaded and unloaded vehicles), alternative routes, etc. Some of the output statistics were waiting time at a certain point, the number of jobs performed by each AGV, and the average time the AGVs have been idle.雪柳黄金缕
Araki et al. (1987) developed a simulator that could handle 68 kinds of AGV path patterns including station, entrance, exit, direct line, curve, T-cross, etc. The simulator consisted of path editor, numerical data input editor, shortest path calculator, simulation executor, animation ction, and results output ction. The shortest path calculator and the simulation executor were programmed i
n FORTRAN, and the rest were programmed in BASIC. The simulation executor had the initial tting, time management, AGV dispatch management, machine tool management, and data management functions.
Schulze and Ronbach (1987) built MATSIM, a special simulation software-tool with a module library, for material flow systems. MATSIM input data included route, vehicle data, priority strategies, and quence of processing. Output data were the total output of the system, number of loaded and empty travels, blocking time, battery charge time, vehicle waiting time, and temporal using of workstations. Ozden (1988) developed a discrete-event simulation program by using LISP to investigate the effect of key factors, such as number of pallets, number of vehicles, and carrying capacity, of multiple-load-carrying AGVs on the overall performance of a FMS. The simulation program provided an animated color view of the FMS operation upon the ur's request. He also stated that the predefined functions of LISP enabled the ur to design simulation models with modularity similar to special-purpo simulation languages.
Mosca et al. (1991) constructed and utilized a transporting network simulator inside a large dimensions fruit and vegetable market utilizing AGVs. The discrete and stochastic simulator was programmed in FORTRAN 77 language. Its general objectives were to plan choices to minimize inv
estment costs, to minimize rvice times, and to maximize AGV utilization in the system. Gaskins and Tanchoco (1989) prented AGVSim2, a discrete-event vehicle system simulator, for evaluating the control strategies of real-time free-ranging vehicle controllers. This simulator was directly linked to the supervisory controller software so that it could be ud to test the intelligent AGV supervisory controller.
>熊蝉
