Performance Analysis of Robotic Cells David P. Harnois MEAE 6960 Modeling and Analysis of Manufacturing Systems Term Project April 12, 2001 Table of Contents Abstract ........................................................................... 3 Introduction ..................................................................... 4 Models ............................................................................. 7 Results ............................................................................ 11 Conclusions ..................................................................... 13 Appendixes ..................................................................... 14 ProModel Models .................................................... 15 Robots ............................................................. 15 People ............................................................. 18 People/Robots Mix .......................................... 21 Simulation Results .................................................. 24 Robots ............................................................. 24 People ............................................................. 28 People/Robots Mix .......................................... 32 References ...................................................................... 36 2 Abstract It is a common practice in today's high tech, fast paced world to implement robotic cells in the manufacturing process. Many companies are able to successfully implement productive robotic cells that decrease process time and reduce product cost. Yet, other companies are not successful in their implementation. negative impact on the manufacturing process. This results in a The goal is to use discrete event simulation techniques to simulate a manufacturing process with both robotic cells and manual labor cells to determine the performance of the robotic cell compared with a manual cell performing the same operation. The belief is there is a better solution than a purely robotic or purely manual cell. Therefore, a third model is introduced that combines elements of a robotic and manual cell together. The process used in this modeling and analysis is the assembly of electric circular saws. Process time was determined randomly, but operations the required more dexterity were shorter when performed by an operator. Assembly steps and times are provided for both robotic and operator assembly. After each type of cell was modeled a new process was modeled that took the fastest process time for each operation, whether it was a robot or operator. This model produced a shorter cycle time than either of the original cells. 3 Introduction This paper addresses the fact that an entirely robotic cell may not be the optimal solution, but looks at a manual and robotic combination solution to achieve the optimum results. Many companies are able to successfully implement productive robotic cells that decrease process time and reduce product cost. Yet there those who are not successful in their implementation. This results in a negative impact on the manufacturing process. Robots work great for simple repetitive process such as installing fasteners, spot welding, and other tasks that only require simple linear movements. They are also appropriate for some complicated tasks such as precision welding and delicate assemblies. However the robot may not be the best choice when an operation has a high failure rate and an operator could easily identify parts the require rework. Robots also may be a poor choice in assembly tasks that require complex movements to manipulate parts for assembly. Robotic cells can also be set up where robots perform all of the actual assembly work, but an operator is responsible for loading the parts and materials for each assembly. This eliminates the need for complex robotics to feed and load the assembly cell. The operator is able to load and monitor the robots, while the robots perform an accurate and repeatable process. Modern simulation tools allow manufacturing engineers to develop manufacturing cells that are built to the optimum layout. If properly used, and if the engineer is willing to consider a multitude of different types of cell layouts the optimum solution can be reached. This optimum solution most likely will not be an entirely automated cell, nor will it be manual. New three Dimensional simulation software is available for modeling robotic cells, in addition to the two dimensional program used in this project. Besides 4 being used to optimize robotic cells, 3-D software allows for offline programming by allowing manufacturing engineers to visualize work cells, identify potential problems, perform feasibility studies, and quickly modify the environment. The models in this paper only reflect changing between an operator and robot to do the actual work. This has greatly simplified the actual considerations that must be made when determining how the work should be divided in and actual cell. In a real life modeling situation, it would be necessary to factor the type of handling, the actual layout, required part set-up, and the operation. The handling aspect relates to the transport of the product from one operation to the next, is it better with a robot or an operator? Factors to be considered when making this choice is distance, queues, and special part characteristics such as temperature or chemical exposure. Cell layout is also important, because the layout may prevent or require the use of certain options. Part set-up can also affect choices. Complicated set-ups may not even be possible with a robot. The type of operation must also be considered, operations that require a lot of dexterity may be difficult for a robot, where simple, repetitive tasks are well suited to a robot. Certain operations may be better suited to a robot for safety issues. When considering all the variable that exist when developing the most efficient cell one can easily see how a company can make wrong decisions. Often times they just assume that robots are the way to go. In some cases cost is not an issue, and robots are the best choice at any cost. The semiconductor industry is a heavy user of robots for both process and handling. The processes used in the manufacture of semiconductors require many toxic chemicals, so robots reduce the risk to operators by keeping them away from chemicals. A second reason that robots are so widely used is the cost of the products. An 5 unprocessed lot of silicon wafers cost can cost over $5000, by the time they reach the end of the process their value exceeds $1 million. Robots reduce handling damage and operator error. In most cases the choice of whether or not to use robots is not a clear cut as in the semiconductor industry and it is important that careful, detailed analysis be made. This paper includes some simple models that illustrate a robotic cell, a manual cell, and a cell that is a combination of robotic and manual operations. By using some simple models and basic assumptions the attempt is to illustrate to the reader that a purely robotic cell may not be the optimum solution. This information may help to prevent expensive robotic lines and cells that are not profitable. 6 Models The system to be evaluated will be a seven-station assembly line for the manufacture of electric circular saws. The system will be modeled in ProModel Student Edition with both a Robotic Assembly Cell and a manual assembly cell. A new system will then be modeled that uses the fastest processing time for each operation to create a cell that integrates the use of robots and operators. Assumptions: It is assumed that there is a dedicated worker or robot at each station. All supplies are delivered to the assembly cell another cell that kits the parts to be assembled. The parts move by automatic transfer station from one operation to the next in both the robotic and manual cells and the transfer time is included in the operation time. Kited parts arrive at the cell every 1.5 minutes via an automated conveyor. The workweek consists of three 8-hour shifts, five days a week, for 120 working hours per a week. Robots fail once every 250 cycles and take 15 minutes to repair. The operators take one 15-minute break every four hours. All products move into the system, through the system, and out of the system by automatic transfer stations. Each Station has a queue of size 25 in front of it. 7 Process: The table below shows the seven different stations and the respective robotic and manual processing times. Process Robot Time Manual Time Attach Electric cord to electric motor and mount motor to bottom half of exterior case. Mount motor to bottom half of exterior case. Attach top half of exterior case. Attach Safety Shield Attach Guide Plate Attach Saw Blade Attach labels and test saw Total Time 1.7 minutes 1.5 minutes 1.1 minutes 1.4 minutes 1.3 minutes 1.5 minutes 1.5 minutes 1.2 minutes 1.7 minutes 1.8 minutes 10.3 minutes 1.3 minutes 1.5 minutes 1.5 minutes 1.2 minutes 9.9 minutes The approach to solving this problem is to model using discrete event simulation and the ProModel software to model a robotic cell, a manual cell, and then modeling a cell that is a combination of robotic and manual. As can be seen from the table some tasks are better performed manually than by robot. In developing this process it was assumed that tasks requiring more dexterity would be performed faster by a human and simple tasks would be performed faster by a robot. By finding the best combination of tasks, the optimal solution can be found and assumptions made as to why robotic cells are not always profitable. The table below shows the process that was modeled using the fastest processing time for each operation. 8 Process Process Time Processor Attach Electric cord to electric motor and mount motor to bottom half of exterior case. Mount motor to bottom half of exterior case. Attach top half of exterior case. Attach Safety Shield Attach Guide Plate Attach Saw Blade Attach labels and test saw Total Time 1.5minutes Operator 1.1 minutes Robot 1.3 minutes Robot 1.3 minutes 1.2 minutes 1.5 minutes 1.2 minutes 9.1 minutes Operator Robot Operator Operator The print outs of the models used in this paper are included in the appendix. The file project-robot.mod is the model of the robotic cell, project-people.mod is the model of the manual cell, and project-mix.mod is the combination cell. These printouts include details on all of the locations, downtimes, entities, processing, and arrivals. There is also a graphical printout for each model. Summaries of all three models are included in the table below. Arrival Frequency Downtime Frequency Downtime Logic Processes: Attach Electric cord to electric motor and mount motor to bottom half of exterior case. Mount motor to bottom half of exterior case. Attach top half of exterior case. Robot 1.5 minutes 250 Operator 1.5 minutes 240 Wait 15 Wait 15 Robot/Operator 1.5 minutes 250 for robots 240 for operators Wait 15 Wait 1.7 Wait 1.5 Wait 1.5 Wait 1.1 Wait 1.4 Wait 1.1 Wait 1.3 Wait 1.5 Wait 1.3 9 Attach Safety Shield Attach Guide Plate Attach Saw Blade Attach labels and test saw Wait 1.5 Wait 1.2 Wait 1.7 Wait 1.8 Wait 1.3 Wait 1.5 Wait 1.5 Wait 1.2 Wait 1.3 Wait 1.2 Wait 1.5 Wait 1.2 10 Results These models showed that there are some advantages to using a hybrid assembly cell that uses both robots and operators. The cell modeled here used faster assembly times for the robot on simple tasks and assigned longer times for tasks requiring more manipulation. The results of this experiment are included in the appendix. The model of the manual cell shows higher throughput than the robotic cell since most operations are require manipulation, hence longer process times. The hybrid cell was essentially the same as the manual cell, therefore there were no improvement in efficiency. In fact, the hybrid cell actually produced slightly fewer saws. The primary reason for this was that the longest processing time was the same in both models. In the hybrid cell, most of the utilizations were lower than in the manual cell. The results are summarized in the following table. Throughput Utilization: Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Step 7 Robot Operator 3868 4502 Robot/Operator Hybrid 4492 96.85 86.03 91.53 94.72 93,12 96.73 96.69 99.98 91.96 99.92 87.91 99.87 99.84 79.84 99.98 67.50 82.50 87.8 75.00 99.63 79.68 Although these models do not show the hybrid cell to be the most optimal, they do show that the robotic cell is not always optimal. The results show that in the 120 hour work week modeled that the hybrid and operator cell both have the same throughput. No noticeable improvement in throughput exists in this case because of the model. In the 11 case of the hybrid cell the longest processing time is 1.5 minutes, which is equal to the longest operation in the operator cell. Therefore the limiting operations in each model are equal, so one would expect equal throughputs. However, upon carefully evaluating the results one will notice that there is more availability in the hybrid cell. A new model was developed that uses a reduced process time for some of the operator steps. This model creates process times that are all shorter than the longest times in either the robot or operator models. The results for these models are shown below. Throughput Utilization: Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Step 7 Robot Operator 1285 1502 Robot/Operator Hybrid 1500 96.81 64.34 80.83 90.42 85.61 96.43 96.33 89.09 98.29 92.35 97.57 99.63 79.63 79.57 99.93 67.50 75.00 82.50 75.00 93.75 75.00 The results of this model still shows the Operator model to have a higher throughput, however when looking at the models it appears that the main reason the hybrid model lags behind is because its throughput is much lower at the first operation between the two models. This fact is difficult to understand since the first process is identical in both operations. It is believe that a bug exists in the software or the model which is causing strange results, since one would expect the throughput at identical operations to be the same when the operation is the very first process. 12 Conclusions This paper illustrates that an entirely robotic cell is not always the optimal solution. The results show that by using a mix of operators and robots better results are obtained than by the use of only robots. The basic models and results presented in this paper show that if a cell is properly modeled, and the use of both robots and operators are taken into account, an optimal solution may exist in which both robots and operators are utilized. It is a rare case where a totally robotic cell is the optimal solution. Besides the straight throughput issue, other factors that must be considered when making these type of decisions is capital costs, repair costs, and salaries. The cost of robots is usually very high, besides capital costs there are spare part and maintenance costs. Many companies have been able to successfully implement productive robotic cells that decrease process time and reduce product cost. These companies most likely took advantage of the tools available today to carefully simulate their processes and determine that robots were the optimal solution. The many other companies who have not been successful in their implementation should take a careful look at their modeling and simulation tools and processes This paper shows how modern simulation tools allow manufacturing engineers to develop manufacturing cells that are built to the optimum layout. The simple simulations used in this paper show that there are a variety of performances based on the makeup of the cell. 13 Appendixes 14