Robotics and Computer Integrated Manufacturing 18 (2002) 241–254 Development of a placement time estimator function for a turret style surface mount placement machine Kimberly P. Ellisa,*, John E. Kobzab, Fernando J. Vittesc a Grado Department of Industrial and Systems Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061-0118, USA b Department of Industrial Engineering, Texas Tech University, Lubbock, TX 79409-3061, USA c i2 Technologies, Dallas, TX 75234, USA Abstract Ef?ciently placing printed circuit board (PCB) components is critically important in the highly competitive electronics industry. Previous research has focused on algorithms to determine feeder arrangement and placement sequencing for placement machines. However, consistent machine models do not exist that can be used to compare the results of these algorithms. A conceptual model of a placement time estimator function is constructed for turret style surface mount technology (SMT) placement machines. Empirical data is then used to develop the estimator function for the Fuji CP4-3 machine. Results for eight commercial PCB designs indicate that there is no statistical difference between the actual and the estimated placement times. Thus, the placement time estimator function developed in this research provides an accurate method to estimate component placements times on a PCB by a Fuji CP4-3. The placement time estimator function presented here can be used as part of a computer integrated manufacturing system to facilitate process planning and improve cycle time estimates for production planning purposes. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Printed circuit board assembly; Process planning; Placement time estimator; Surface mount placement machine 1. Introduction The electronics industry continues to rank as an important industry as the twenty-?rst century begins. Manufacturers strive to provide electronic products with advanced technology while meeting demands for increased productivity, higher quality standards, improved customer responsiveness, and cost effectiveness. Most electronic products manufactured today contain printed circuit boards (PCBs) as critical elements and the PCB market currently generates worldwide annual sales of $35 billion [1]. The assembly of PCBs is a complex task involving the placement of hundreds of electronic components in different shapes and sizes at speci?c locations on the board. Surface Mount Technology (SMT) has become the choice of manufacturers due to the high precision of surface mount equipment. SMT equipment, however, is expensive with each machine ranging in price from $250,000 to $1,000,000. Manufacturers often have several assembly lines in a production facility, and an assembly line generally contains multiple SMT placement machines. In order to remain competitive in the PCB market, manufacturers must concentrate their efforts on improving the ef?ciency of their SMT assembly lines. Production planning and control, process planning, and quality control are important activities for achieving this ef?ciency in the PCB industry. Of these activities, process planning is particularly important due to its direct impact on ef?cient and responsive printed circuit board assembly operations. The purpose of this paper is to develop a detailed operational model of a turret style surface mount technology placement machine. These machines, commonly referred to as chip shooter machines, are prevalent in industrial facilities and are manufactured by Fuji, Panasonic, and Universal equipment providers. This paper presents a placement time estimator function to predict the placement time to populate a board for a given design and a given process plan. This estimator function provides a means to evaluate and compare process plans for these machines in an effort to improve the overall productivity of an SMT system. Following the development of the placement time estimator function, an experiment using a Fuji CP4-3 Chip Shooter machine is conducted to determine the relevant machine parameters. The new placement time estimator function is then validated by comparing the estimated placement times to the actual placement times for a set of PCBs. 2. Background In printed circuit board assembly, process planning involves two closely related issues: setup management and process optimization. Setup management involves: * assigning PCBs to the different SMT lines, grouping PCBs into families, grouping placement machines, and * sequencing the production of PCBs. These issues are addressed to reduce setup times, balance capacity across multiple SMT lines, and reduce inventory levels. Process optimization involves: * allocating components to the placement machines, and * configuring a machine for operation. The machine con?guration issues may include feeder assignment and placement sequencing decisions depending on the characteristics of the placement machines. Process optimization is performed to balance the workload across machines in an assembly line and reduce component placement time for the machines in the line [2–4]. A variety of researchers have investigated process planning issues for turret style placement machines. McGinnis et al. [4] describe the relationships between the feeder assignment and placement sequencing problems for both sequentialand concurrent type machines. Sequentialmachines retrieve, position, and place components in a sequential patter. Concurrent machines perform two or more of the operations simultaneously as with the chip shooter machine. The authors also present a general framework for estimating the placement time for both sequential and concurrent machines. In other research, the focus has been on the machine con?guration issues of feeder arrangement and placement sequencing. Moyer and Gupta [5] present a solution approach for determining both the feeder assignment and placement sequence for a chip shooter machine. In this paper, the authors assume that the chip shooter machine operates in more of an asynchronous mode than a concurrent mode. Leipala and Nevalainen [6] develop a feeder arrangement and placement sequencing approach for a Panasert RH machine, which is a turret type machine with two placement heads. Although the machine operates in a concurrent manner, the authors do not appear to incorporate the performance measure of the solution approach does not appear to incorporate this. Leu et al. [7] present a solution approach for the component sequencing and feeder arrangement problem for three types of placement machines, including a chip shooter machine. They present a model of a turret style placement machine with two heads on the turret, a PCB table with constant velocity, and a feeder carriage with constant velocity. De Souza and Lijun [8] develop a knowledge-based approach to address the feeder arrangement and component sequencing problem for the chip shooter machine. The objective function used to evaluate the approach, however, is not presented. Bard et al. [9] develop an approach for feeder arrangement and placement sequencing for a Fuji CP-II machine. In their model, however, the PCB table and the feeder carriage may begin movement asynchronously. While a wide variety of research has been conducted on optimizing the performance of the prevalent turret style placement machine, a unifying description and detailed model of the operational characteristics of the machine is not available. Thus, it is dif?cult to compare or validate research results. This paper presents a detailed model, using empiricalresults, to provide a common baseline for research efforts and comparison in this area. 3. Placement time estimator function In order to evaluate a potential process plan, it is necessary to measure or estimate the resulting machine placement time. In this research, a placement time estimator function for the turret style placement machine is developed to evaluate a process plan. The estimator function accounts for the operational characteristics of the chip shooter machine (such as concurrent mechanisms, the PCB table speed, the turret head speed, and the feeder carriage speed) as well as the characteristics of the components that are mounted. Following the development of the placement time estimator function, an experiment using a Fuji CP4-3 Chip Shooter machine is conducted to determine the relevant machine parameters. The new placement time estimator function is then validated by comparing to the actual placement times for a set of PCBs. Some of this description also appeared in Ellis et al. [10], but the focus in that article was on developing a process plan rather than on the detailed operation of the surface mount machine. As illustrated in Fig. 1, the chip shooter machines operate in the following manner: a component is retrieved at the grip station from a feeder mounted in a pre-speci?ed slot on the feeder carriage, while the placement station simultaneously mounts a component at a pre-speci?ed location on the PCB. The time for the grip station to pick a component from the positioned feeder carriage and the time for the placement station to place a component on the positioned PCB are ?xed and equivalent. This time is referred to as the ?xed pick and place time (FPP). After these operations are completed, all the mechanisms move simultaneously to the next position. The turret rotates one position, the PCB table moves to the coordinates of the next component in the placement sequence, and the feeder carriage moves to position the feeder of the appropriate component (the component that is NH=2 placements in the future) under the grip station. The parameter NH represents the number of placement heads on the turret. This cycle repeats until all the components have been mounted on the board. The following sections describe the PCB table movement time, the turret rotation time, the feeder carriage movement time, and the ?xed pick and place time. Fig. 1. Example of a turret style placement machine. 3.1. PCB table movement time The PCB table movement time is determined by the maximum movement time in the x-direction and y-direction. Since the table moves simultaneously and independently in the x and y-directions, the Chebychev distance metric [11] is used. The table movement time for the ?rst component, however, is zero since this movement is performed while the PCB is loaded in the machine. The PCB table movement time is represented as follows: In this expression, x?i-1問 and x?i問 represent the x-coordinates for component placements i -1 and i in the sequence, y?i-1問 and y?i問 represent the y-coordinates for the component placements i -1 and i in the sequence, and NC represents the total number of component placements. The PCB table has multiple speed settings. Each component has a speci?ed table speed setting, STS?i問; representing the fastest table speed allowed when the ith component is placed on the board. Some components (such as larger and heavier components) have a slower table speed setting to minimize the possibility of them moving out of position. The functions VX and VY represent the PCB table velocities in the x and y-directions in terms of the distance traveled and the current table speed setting s?i問: The PCB table accelerates from its initial point of movement, reaches a maximum velocity, and then decelerates to a stop at its destination point. Thus the average velocity is higher when traveling longer distances than for shorter distances for any table speed setting. The value of s is zero for the fastest table speed and N for the slowest table speed setting. The speed setting for the placement of a speci?c component is dictated by the slowest speci?ed table speed setting of the components already mounted on the PCB, Once the table speed setting is reduced, it stays at that setting unless it is reduced again. Thus, the table speed setting can decrease but never increase while placing components on a given PCB. 3.2. Turret rotation time The turret head mounts a speci?c component on the PCB while it simultaneously retrieves a component from the feeder carriage that will be placed in NH=2 placements. As the turret rotates to the placement location, the components on the turret head are rotated into position and also inspected using automated vision inspection. Components are retrieved from the feeder by a suction nozzle. Since some components are more dif?cult to hold by suction (such as larger, heavier components), they must move slower. Components of different sizes are mounted using different turret rotation rates, with larger components typically having slower rotation rates. The turret head rotation time is dictated by the component with the slowest turret rotation rate loaded in the turret. Thus, the turret rotation time associated with the ith component placement, ttr?i問; is the maximum of the turret rotation times associated with the NH=2 components loaded on the turret. The turret rotation time associated with the ?rst component placement is zero since the turret movement for the ?rst component occurs while the PCB is loaded on the machine. The turret rotation function is represented as follows: The turret has a minimum turret rotation time, MRT, achieved using a 100% turret rotation percentage. Each component has a speci?cation turret rotation percentage, STR?i問; associated with the ith component in the sequence. For the last NH/2-1 component placements, there will be some empty turret heads since all the components from the placement have been picked up from the feeder carriage. Since the speci?cation turret rotation rate for empty heads is 100%, : 3.3. Feeder carriage movement time The feeder carriage movement time is determined by the movement time between feeder slots in the x-direction. Let FS?i問 represent the feeder slot associated with the ith component placement. The feeder carriage movement time associated with the ith component placement, tfc?i問; is the traveling time from the feeder slot of the ith t NH=2 -1 component placement in the sequence to the feeder slot of the ith t NH=2 component placement in the sequence. The feeder carriage movement time for the ?rst component, however, is zero since this movement is performed while the PCB is loaded in the machine. In addition, the feeder movement time associated with the last NH=2 components is also zero since all the components to be mounted have already been loaded on the turret while mounting the previous components. Thus, the feeder carriage does not move after the NC -NH=2 component placement and FS?j問? FS?NC問 for j > NCFNH=2: The feeder carriage movement time is represented as follows: > 0 for i ? 1; < tfc?i問? > ejFS?itNH=2-1問-FS?itNH=2問jT for i ? 2to NC: e3T : VfceFS?itNH=2-1問; FS?itNH=2問T Note that Vfc represents the velocity function for the feeder carriage mechanism as a function of the number of feeder slots traveled. The feeder carriage functions similarly to the PCB table in that it also accelerates from an initial feeder slot position and decelerates to a stop when reaching the destination feeder slot. Thus, a higher average velocity is obtained when traveling many feeder slots versus a few feeder slots. K.P. Ellis et al. / Robotics and Computer Integrated Manufacturing 18 (2002) 241–254 3.4. Fixed pick and place time The pick and place time required to retrieve a component from the feeder carriage and simultaneously mount a component on the PCB after all the mechanisms are positioned is ?xed and referred to as the ?xed pick and place time (FPP). 3.5. Placement time function for a sequence of components Due to the concurrent operations of the placement machine, the placement time of a component is determined as the maximum of the movement time of the turret, the PCB table, and the feeder carriage, plus the ?xed pick and place time such that: PT?i問? maxfttb?i問; ttr?i問; tfc?i問gt FPP: e4T The total placement time for placing a total of NC components is the sum of the placement times associated with each of the components such that: NC PT ? X max max MRT ; max jx?i問-x?i-1問j ; jy?i問-y?i-1問j ; ejFS?itNH=2-1問-FS?itNH=2問jT i?2 ipjpitNH=2-1 STR?j問 VX ex?i問; x?i-1問; s?i問TVY ey?i問;y?i-1問; s?i問T VfceFS?itNH=2-1問; FS?itNH=2問 NC X t FPP; e5T i?1 where s?i問? max fSTS?j問g: 1pjoi As the PCB is loaded onto the machine, the ?rst NH=2 components are loaded on the turret, the table moves to the location of the ?rst component, and the feeder carriage moves to the pickup location of the eNH=2Tt 1 component in the sequence. Thus, the placement time of the ?rst component is the ?xed pick and place time (FPP) since none of the three mechanisms are required to move for the ?rst placement. In addition, recall that for the last NH=2 components in the placement sequence, the feeder carriage movement time is zero since there are no more components to retrieve, so FS?i問? FS?NC問 and tfc?i問? 0 for i > NC -NH=2: The speci?cation turret rotation percentage associated with an empty head is 100%. The notation used to develop the placement time estimator function is summarized in Table 1 in the Appendix. 4. Empirical results for placement time function The values of the parameters of the placement time function can be obtained from data provided by equipment manufacturers or from experimental analysis. Equipment manufacturers are hesitant to release this information due to the competitiveness of the PCB assembly industry. As part of this research effort, experiments were conducted using a Fuji CP-4 placement machine to determine the following parameters: * minimum turret rotation time; * ?xed pick and place time; Table 1 MRT and FPP experimentaldata Turret rotation (%) No. of components Placement time (s) 100 40 7.11 100 40 7.09 100 40 7.07 100 40 7.12 50 40 12.94 50 40 12.90 50 40 12.88 50 40 12.86 K.P. Ellis et al. / Robotics and Computer Integrated Manufacturing 18 (2002) 241–254 * PCB table average velocity functions for different table speed settings; and * feeder carriage average velocity function. This section describes the results of the experiments for the Fuji CP4-3 placement machine. The values of the parameters obtained from the experiments are then incorporated into the placement time estimator function. The resulting placement time function provides the information necessary to estimate the time to place the components on a printed circuit board for a given feeder arrangement and placement sequence. 4.1. Calculation of minimum turret rotation time and ?xed pick and place time An experiment was conducted to determine the minimum rotation time (MRT ) and the ?xed pick and place time (FPP) for the Fuji CP4-3 machine. The experiment was conducted using two different speci?cation turret rotation percentages eSTR ?50% and STR ?100%) with four replications. The experiment consisted of placing NC=40 components and measuring the placement time. The components were retrieved from the same feeder and the fastest table speed setting was used with a placement distance between consecutive components of 5 mm so that the turret rotation time would be longer than the PCB table movement time and the feeder carriage movement time. Thus, the total placement time is the sum of the turret rotation times and the ?xed pick and place times associated with each component placement. Note that for the Fuji CP4-3 machine, there are 12 heads on the turret device. Table 1 summarizes the data obtained from this experiment. As described previously in Eq. (5), the number of rotations that the turret performed is NC -1 ?39 since the ?rst component placed is already located at the placement station when the placement of components starts. The number of ?xed pick and place movements is NC ?40 sin