Final Report For Responsive Automotive Manufacturing Plant
Download 1.02 Mb.
|
1 2
- Navigate this page:
- Process Rows in the matrix
- Pallet Inspection
- Phased introduction of equipment
- Proposal Partners
- Steering Group Members
- Presented to
ModelIn the model, empty pallets entered the facility in a random mix, at a rate computed to produce the predicted annual volume in a working year, e.g. for Year 6 the total number of vehicles required is 103600 (from Table 1 for Year 6). On entering the conveyor they were configured for the particular variants required. As the pallets travelled along the conveyor belt they were populated with components that in reality would have been loaded from racks. These racks were not shown in the model but could be added along with the operators if required. This would not affect the model output, merely provide more work for the computer to carry out. However, it is realised that in reality, the loading operation would present a large logistics problem to be solved. The facility was configured to separate the medium wheelbase body sides (MWB in Table 1) from the short wheelbase body sides (SWB in Table 1), so that for Year 6 the first row of process machines (welding) were organised as follows: seven machines to the right of the facility were dedicated to welding MWB (green components) two machines to the left were dedicated to welding SWB (yellow components) the remaining two machines (third and fourth from the left) could weld either body side. The predicted SWB production for Year 6 would require only three welding machines for a constant pallet input rate. However, because the input was random there were certain times when large numbers of SWBs emerge without interruption and this was catered for by the extra shared welding machine. Process Rows in the matrixWhen the pallets were fully populated, they left the conveyor and were carried by automated guided vehicles (AGVs) to the first row of process machines (in this case welding machines). The assemblies stayed at the welding machines for a specified time appropriate to the variant being processed. A welding machine was chosen to conduct the process based on the following logic: Is the pallet carrying SWB or MWB? Is there an assembly waiting to be taken from the current process to the next process, and the machine not being maintained? Is there a machine without an assembly waiting to be processed, and the machine not being maintained? Is there an input buffer without an assembly on it, and the machine not being maintained? If all machines are occupied or being maintained and all the input buffers have an assembly; wait until an assembly needs moving to the next process. This logic was applied to machines sequentially. For those machines capable of processing MWBs (nine machines to the to the right) the machine furthest to the right was the first to be considered and the last to be considered was the ninth one from the right (capable of processing both SWB and MWB). For those machines capable of processing SWBs (four machines to the left) the machine furthest to the left was the first to be considered and the last to be considered was the fourth one from the left (capable of processing both SWB and MWB). When a suitable machine had been selected, an attribute of that machine was set to prevent another assembly going to it. This attribute is re-set when processing had been completed. After leaving the welding process, the AGVs moved the assemblies to the next row in the matrix, in this example riveting. As before some machines were capable of riveting MWBs (four), some were capable of riveting SWBs (two) and one was shared. Again a machine was chosen using the logic described above and again each assembly had a different time specified for it’s processing. There was sufficient AGV track length between the welding and riveting processes to act as a buffer and accommodate any AGVs waiting to enter the riveting stage. A queue can form with this technology because the input rate is not constant, but there are sufficient machines to process the specified number of assemblies over the specified time. After riveting the assemblies were moved by AGVs to the next row in the matrix, where they were adhesively bonded. There were only two bonding stations in this example and so sending alternate assemblies to each machine would be the simplest logic to apply and this would work if the bonding process times were similar for each variant. However, the same logic previously used was applied so that the number of bonding machines could be increased if required. In reality the cells could perform any process required within a specific facility and each cell could be either dedicated to one model or capable of producing all the required models. After this the assemblies were carried to a framing station for further welding, a frame riveting station and finally to a frame bonding station. When all processes had been completed, the assembly was taken to the stockpile; for the simulation the frames disappear and so reduced the computer’s workload. PrototypesIt was thought desirable to have the capability of being able to follow and observe a prototype being processed throughout the matrix. Consequently, prototypes follow a defined path through the matrix and are the only assemblies that do not follow the logic sequence to select a process machine. However, they have an affect on the logic sequence by occupying a machine and so denying it to other assemblies. Prototypes were directed to particular machines at each stage and were then carried to an prototype inspection area. In this way all the processing could be observed and inspection carried out after each process has been completed. Pallet InspectionThe geometry of the empty pallets was checked in an inspection area at the side of each return track. Pallets should be designed to accommodate all build types; in this example twelve build types (five SWBs and seven MWBs from Table 1). As the pallets entered the matrix, each build type was assigned an attribute that was consecutively number up to one hundred and then started from one and continued to repeat this cycle. This provided an easy method of removing a percentage for inspection as removing all those with numbers greater than ninety, would remove ten percent and all those with numbers greater than eighty-five, would remove fifteen percent etc. Phased introduction of equipmentIn this example, the production equipment required to meet the predicted annual production numbers, increases every year until year 6 and then decreases. Using the RAMP method, equipment is purchased as required and so the amount of equipment in the facility increases until Year 6 then decreases, and this can be seen in the following figures 2 & 3. As can be seen, the equipment required initially is not great as the numbers to be produced have yet to reach their highest levels. Once demand decreases the number of machine again falls. However, in a real plant these machines would more likely be switched to the production of new vehicles. The simulation was created using robots for the processes at each cell but this could easily be human operators or other types of equipment. Robots are shown for ease of simulation.  Figure 2. RAMP Year 1  Figure 3. RAMP Year 6 Cost ModelAn outline cost model has been developed for the project, in the timeframe of a feasibility programme it would be impossible to construct a detailed cost model. It was intended to use a high level generic model to show the feasibility of the RAMP concept. However, the data does not exist that can show at a generic level where the RAMP paradigm is effective. In many cases the cost data is unavailable or companies do not know the detailed costing for any particular vehicle type. The model has therefore been designed as a high level ‘question and answer’ structure such that any individual company could tailor it to put in their actual high-level cost data and determine whether RAMP would be feasible for them. The model can then be added to, giving a ‘drill-down’ function where more detailed costs are available and/or need to be investigated. An initial literature search showed that although there appear to be many micro-models and discussions of how to assign costs to particular manufacturing areas there appears to be no overall models publicly available showing the overall costs of vehicle manufacture. Published industry statistics were analysed to see what this might reveal about the cost equation. The Society of Motor Manufacturers and Traders (SMMT) publish monthly statistics showing the numbers of vehicles produced by each manufacturer. The reports give details of exports and new vehicle registrations as well but for simplification the initial start point was the actual production figures. The annual summary SMMT report gives an overview of the industry and annual production figures for each individual model. Taking the annual figures for individual models these were plotted yearly from 1979 or the start of the model life to date. Initially this was done for Rover group as a benchmark. The results were surprising since the automotive industry prides itself on its high capacity utilisation. The justification for single-line factories and high volume manufacture is the efficiency of production that leads to high equipment utilisation. However, the graphs show that consistently high levels of production are not achieved and certainly not throughout the model life-cycle (see Figure 4). The Rover 800 graph for example clearly shows a peak production rate that quickly drops off. (see Figure 5).  Figure 4. Total Land Rover Capacity Utilisation  Figure 5. Rover 800 Production Similar graphs were plotted for other manufacturers to see if this was peculiar to one company. Surprisingly the results were similar for every manufacturer sampled, including Toyota and Nissan. It seems that manufacturers can hit high utilisation for a limited period but the ramp-up and down time can be significant. T  he graphs do not include sales data and this of course may have a great impact on production if a new model fails to take off. Figure 6. Ford Fiesta Production Given that even very successful models like the Ford Fiesta (see Figure 6) do not have sustained periods of very high volumes suggests that the manufacturing facilities may not be being best utilised. Also industry trends show that model lifetimes are shortening and therefore the payback period for any dedicated equipment will be shorter. Harbour report 1994 ‘Number & types of model have increased exponentially while the average volume per model is much less’ ‘models change frequently, often every 4 years’ ‘The plant that can cost effectively produce several different models and change from an old model to a new one quickly will have the competitive advantage’ Thus the results from using the SMMT data showed that the RAMP concept might be applicable across a higher level of volumes than initially thought. There are no available statistics to show sales compared with capacity or indeed demand as in some cases sales will have been heavily discounted in order to move stock and in others demand may exceed supply. The RAMP concept assists in both these scenarios, as the facility will be able to flex in line with demand – although clearly the total factory capacity cannot be exceeded. However, owning a number of RAMP facilities as opposed to dedicated lines will alleviate the above scenarios in most cases. From an initial plot of production figures therefore it seems likely that RAMP is a feasible concept. Looking at the industry benchmark figures produced by the ‘Harbour Report’ (these are only available publicly for US companies but provide a useful comparator and are used by UK industry for benchmarking). The capacity utilisation figures do not seem to stack up with the figures from the SMMT as the Harbour Report shows consistently high capacity utilisation. However, these figures often exclude plants during change-over to a new model. Plotting figures from the Harbour Report it is difficult to compare statistics since the figures quoted are for plant capacity utilisation and this can be affected by a number of things such as different shift patterns and overtime working. Also, for the years studied (from 1984 – 2000) the Truck industry in particular in the US was recording record sales and many factories were moved over to truck production. Measures quoted are often Hours per vehicle or workers per vehicle, which hide a multitude of issues and do not specifically address profitability. The best production line in the world will not produce a profit if the cars do not sell at a price sufficient to more than cover the costs e.g. Chrysler in 1994 made twice as much money on a profit-per-unit basis than Ford although Ford ran at maximum capacity. Nissan was top or close to top in terms of capacity utilisation and HPV throughout the time from 1984 – 2000 and yet consistently failed to make a profit and had to be rescued by Renault in 1999. Predicting what will sell and in what numbers is an almost impossible task so under or over production and corresponding discounting schemes are common. For example in 1994 in the USA there was a huge demand for trucks and not enough vehicles being produced to meet it. Increasing capacity quickly is difficult, either huge overtime costs are incurred in terms of cost and workforce strain or further capacity needs to be introduced by adding another line. This is risky as the line itself will be a big cost and take time to construct, by which point demand may have already dropped off due to waiting time and / or competitor activity. Having a manufacturing facility flexible enough to change the type and numbers of models produced such that customer demand is met would greatly improve profitability. However, it is recognised that capacity may also be severely limited by supply of parts from suppliers as fluctuating demand may not be something they are equipped to deal with. This may be a good reason for implementing RAMP in Tier 1 suppliers first. This would benefit Tier 1 suppliers as they are already producing parts for several different VM’s in parallel. Introducing RAMP would not require any general change in strategy. It is also recognised that whilst a more frequent introduction of new models, if this can be done with minimal disruption to current production, may prove profitable, the cost of design and development must be kept as low as possible. There will be a limit to how frequently models can be changed whilst attracting sufficient buyers to cover development costs and make a significant profit. However, it may also be impossible to know exactly where this limit will lie for individual manufacturers since it is doubtful that any specific market research has been done in this area. The starting point for the model is that a multitude of factors affect cost at a factory level Design costs Component costs Capacity utilisation Facility cost Recurring costs Level of automation Labour
Changeover Maintenance If known, these figures can be put into the model along with similar figures for a RAMP facility. Whilst it is predicted that initially a RAMP facility may cost 20 – 40% more than a traditional line the real benefits would be seen only when the facility is manufacturing a number of different models. Therefore with each new model introduction and therefore change of equipment usage a RAMP facility would benefit from needing only a limited amount of spend to change equipment over rather than install a whole new line and would therefore become progressively cheaper as an option. The following graph, figure 7, indicates how this should be achieved (the actual figures are an estimate based on the cost of a new plant for a 100k per year volume. F  igure 6. Cost of facility (£ Million) The model has been developed as far as it can be with hypothetical cost and volume information. The next phase would be to input data for a real project and compare against a traditional system. SeminarThe RAMP project ran a seminar in July 2001. A number of senior managers were invited and the RAMP team presented the concept and the models constructed so far. Although the number of attendees was not high, the project concentrated on ensuring that the relevant people from each company attended. This ensured that the concept was presented to the most appropriate people. A number of presentations were also given to individual companies. Appendix A lists the companies involved in the Steering Group the companies who attended the seminar other companies attending presentations / visits to discuss RAMP companies contacted about the RAMP project. Proposal & PartnersIn order to spread the ideas to a wider audience , as well as the seminar held in July 2001, several articles have been published in the popular automotive press. The published articles are listed in Appendix B. It is also intended to publish the results more widely over the next few months. The models have been completed as far as they can be with generic data. In order to take the concept to the next stage, ‘real’ data is required. Thus, until industrial partners are ready to make this bold step it is not possible to complete a proposal for further work. The concept is still being disseminated widely to people at the appropriate level within automotive companies. It is to be hoped that a visionary company will agree to become a fully participating partner in the near future. Appendix A Lists of Companies Steering Group MembersMike Shergold Land Rover Richard Hewitt Comau Jason Rowe Lotus Bob Mustard Stalcom Tony King Mayflower Stephen Buckley Mayflower Seminar AttendeesMike Shergold Land Rover Richard Hewitt Comau Ian Oswell Meritor Anna Kochan World Automotive Manufacturing John Rutherford Ove Arup John Burtin Ove Arup Robert Williams Pera Joe Miemczyk University of Bath - 3-Day Car Ashley Roberts DTi Bob Mustard Stalcom Craig Turner Stalcom Robert McQuire Lotus Robert Wapshott Lotus
Presented toMartin Cunliffe Jaguar Cars Nick Fry Prodrive Alaa El-Shanti Bentley Toby Proctor Automotive Online Dean Palmer Manufacturing Computer Solutions Richard Quinn Visteon Terry Cooper Comau Mickey Howard University of Bath, 3-Day car Nina Young Mayflower Kevin Millage Mayflower Contacted: Ken Giles Aston Martin Andrew Mueller Lander Automotive Ltd. Jim Adams DANA Spicer Axle, Europe Alan Seeds British Steel Paul Markwick Ricardo Test Automation Ltd. Appendix B Published Articles The following articles have been published about the RAMP project to disseminate the ideas more widely: ‘The UK’s future lies in niche model production’ - World Automotive Manufacturing – August 2001 RAMP Project - In Automotive Online – October 2001, written by Toby Procter ‘A Philosophy to redefine car production’ – automation, October 2001 in the Industry News section. Page . Directory: fac fac -> Dsci 3870: Management Science fac -> Handling Indivisibilities fac -> Building: 300 Subjects/Programs to be Housed fac -> Plo timeline fac -> Personal Information Name: Maha Al-Ammari Nationality: Saudi Relationship Status fac -> Course overview fac -> Curriculum vitae wei chen professor fac -> Digital image warping fac -> Samples of Elements Exam Question III contains All Prior Exam Qs III except Download 1.02 Mb. Share with your friends:
|
1 2