Boeing QCP Project



|Boeing QCP Project |

|University of Idaho – Quality Management |

| |

|The following is the analysis of the spar quality control process currently in place |

|at the Frederickson, WA plant for the Boeing Company. |

| |

|Dustin Smith, Ian Chestnut, Shane Wemhoff, Nic Pentzer and Jeremy Wemple. |

|12/17/2008 |

| |

Contents

EXECUTIVE SUMMARY 3

QUALITY CONTROL POLICY IN AN OPERATIONAL CONTEXT 4

PRODUCT CHARACTERIZATION 5

PROCESS FLOW CHART 6

CURRENT INSPECTION PROCESS 6

CURRENT PROCESS STATISTICS 6

CURRENT THROUGHPUT & COSTS 7

NEW QUALITY CONTROL POLICY 7

NECESSITY OF MEASUREMENTS 8

MEASUREMENT SYSTEMS 9

MEASUREMENT LOCATIONS 12

TIME OF INSPECTION 13

INSPECTORS 13

SAMPLE SIZE 14

MEASUREMENT FREQUENCY 15

QUALITY CONTROL RESPONSE TO MEASUREMENTS 19

DATA DISPLAYS 19

DATA ACCESS 20

QUALITY CONTROL GOALS & CONCLUSION 20

APPENDECIES 21

FIGURE A 21

FIGURE B 22

FIGURE C 23

WORKS CITED 24

EXECUTIVE SUMMARY

Nice work, tables and figures should be in the body next to where they are discussed, not in the appendix. Why did you leave the word vacuum in, aren’t all hold downs clamps? 98%

The following is the analysis of the spar quality control process currently in place at the Frederickson, WA plant for the Boeing Company. The following was compiled by Ian Chestnut, Nic Pentzer, Dustin Smith, Shane Wemhoff, and Jeremy Wemple from the University of Idaho.

Background and Purpose

The project was given to our team by Randy Schimon, Rick Jones, and Thad Retzlaff. The current quality control process results in a lack of throughput. Also, there is no formal criteria for implementing NC program changes and no formal reporting system to upper management.

Scope

We were asked to analyze the current quality control process in place and formulate recommendations that would result in an overall improvement in the inspection process, the ability to establish criteria for implementing NC program changes.

Findings

We found the current policy results in the potential for traveling defective parts, large amount of inspection hours, it doesn’t effectively monitor variation in the mills, and it doesn’t assign cause to the variation detected by the inspection.

Recommendations

We recommend 2 different inspection processes that will ensure the NC program is correct and the mill is constantly monitored. After an NC program change, completely inspect 3 parts to ensure NC program accuracy. Once the NC program has been accepted, move into a second inspection phase. Every milled part is inspected at randomly sampled locations. This ensures the mill is monitored for variation and all parts are inspected for defects. Furthermore, NC program changes should be implemented in packages, allowing for minimal impact to the inspection process.

QUALITY CONTROL POLICY IN AN OPERATIONAL CONTEXT

While looking at our Boeing project we realized there are two main objectives we are really trying to accomplish. First of all, we want to provide Boeing with answers to their deliverables they presented to us. They took the time and initiative to provide us students with a valuable project and we can repay them by putting forth our best effort to provide possible solutions for them. These deliverables were included in the presentation we delivered on December 5th, 2008. Secondly, we are working on this project to complete the class requirement and achieve the best grade we can applying our acquired knowledge of quality management. We expect this to be a learning project first and foremost. Secondary expectations we have in mind include creating a quality control process that Boeing will actually implement or implement parts of, receiving good grades for our efforts, and using this project for networking in future job searches.

We took this mindset into the creation of the following quality control policy (QCP). The operational context of our QCP is to be applied to Boeing’s Skin and Spar facility in Frederickson, Washington. Specifically we are focusing on the first four steps of the spar and stringer production process. These are the NC program phase, raw material inputs, machine milling process, and the dimension inspection phase. After carefully analyzing all of these phases and applying quantitative and qualitative research we constructed this QCP.

Our QCP will fit into Boeing’s organizational culture because it takes the assumptions that everyone within the plant must be on board in order for our QCP to work. Employees will retain their same position in the company and the current quality control manager would be put in charge of this new process. Employees would just be asked to implement this policy to improve process production quality and process speed.

We do not believe that Boeing’s core business rules will be affected by the changes our QCP suggests. It will allow for a higher throughput rate and higher quality product out the door faster. The only question we have currently is whether the Federal Aviation Administration will approve of our new sampling method. Currently, any measuring or sampling method used by Boeing is reviewed and either accepted or rejected by the FAA. This could possibly limit the effectiveness of our QCP design if our method gets rejected by the FAA. The industry is in the predicament of selling planes to airlines which currently have no money. Our QCP would allow Boeing to become more efficient and effective in their stringer and spar plant. Increasing efficiency and effectiveness can only help Boeing and allow them to focus on other parts of their process that needs attention. This would help them increase production in the stringer and spar plant and allow them to stay on schedule and keep up with airplane production.

The overall economy is currently in a downturn. In particular Boeing is facing stiff competition from its main competitor, Airbus. While we haven’t attempted to solve the competition problem, an increase in production speed and high quality throughput can only help. Whether it means getting more planes to the market faster, freeing up time to work on other things, or saving Boeing thousands of dollars in scrapped material, we feel that our QCP will have a positive impact, even if some very small way.

PRODUCT CHARACTERIZATION

The correct product in the process we are analyzing are stringers and spars of the correct engineering dimensions, correct material from suppliers, minimal duration time of total process, and minimum average resource usage (employee hours, milling time) per part (lineal foot). Boeing produces different sizes and shapes of stringers and spars to fit their different aircraft sizes. The dimensions of the stringers and spars are designed in the engineering department and then brought to the production floor where they are entered into an NC program which communicates with the high speed milling machines essentially telling the machine how much and where to cut on each stringer or spar. Our QCP must ensure that the designs from engineering are accurately entered into a NC program and that the NC program accurately does its job of telling the machine what to mill. The mill needs to be running without hiccups so that it can cut accurately to spec and the inspection phase should inspect accuracy of the process by directly measuring off of the engineering design papers.

Our QCP attempts to minimize the total process time by increasing the number of stringers and spars measured to 100% while at the same time decreasing sample size per stringer and spar. Our dramatic decrease in sampling size will allow a normal inspection reduction time of roughly 6 man hours per 90 foot stringer to about 25 minutes per 90 foot stringer. This will be discussed in further detail later on in the QCP. Milling time is based upon the complexity of the engineering designs and the length of the part being milled. One way to improve here would be purchasing faster machines. However, Boeing recently purchased new high speed milling machines so an equipment upgrade isn’t necessary at this time, as they already have the latest technology. Therefore, to gain efficiency and throughput, Boeing would have to either start recording data real time off the mill (Which would be best but no measuring system is currently in place) or from the inspection phase. They could then turn this data into control charts on a computer and easily monitor the process to determine trends and see when the process starts to get out of control. This is a preventative measure looking at fixing the problem before it causes parts to be out of spec.

Employee efficiency also needs to be examined to see if employee hours can be reduced (therefore cutting costs) due to increased efficiency and effectiveness of the process. Our QCP addresses this by decreasing sample size per part or decreasing the number of places on a part which have to be measured. Here we could drastically reduce the total number of inspection hours and therefore eliminate some need of employees in the dimension inspection phase of the process. Again this is addressed in more detail later on in the QCP.

PROCESS FLOW CHART

We have developed a process flow chart to help show how our new quality control policy works. A graphical representation can be found in the appendices FIGURE C. tables and figures should be in the body next to where they are discussed, not in the appendix

CURRENT INSPECTION PROCESS

Boeing currently does full inspections on parts and designates them as either a pass or a fail. This is not to say that every part is inspected. Depending on what Boeing calls an “inspection level”, some parts are skipped and not inspected at all. This method is commonly known as Pre-Control Limits or Narrow-Limit Gauging. Currently, this method is enacted to determine whether or not an NC Program change is correct while in reality it is catching mill variation. There are drawbacks to this method and they are somewhat significant:

• No Control Charts

• You cannot determine any patterns that occur within your samples because you cannot visually see them, this also eliminates any diagnostic techniques you could do to fix patterns drifting out of spec and lastly, you lose any history of the process for benchmarking.

• Small Sample Size

• Can’t detect mean shift

This method of sampling works best when based off of processes which have a capability index much greater than one (recommended of at least 2). Moreover, you still cannot detect mean shift and in this case when an inspection fails the part can still be within spec. Lastly, this process misses a lot of parts and in theory is missing a lot of failed parts. This is partly because of the way this sampling method works; it is a random inspection of the entire part at a random interval until it reaches level 0 when it becomes a full inspection of every part. Inspections at level 1 could theoretically miss up to 18 parts in a row.

CURRENT PROCESS STATISTICS

Currently, according to our data of 2,666 entries over 18 months, 66 had failed while 2600 were either skipped or passed, refer to appendices FIGURE A. The figure also supports the theory that when one part fails, you will have many failures happen. This also shows a failure rate of 3%. Currently, when Boeing is at level 0, they are inspecting 30 parts, the book recommends sampling only 25, and this change increases our odds to make it through without finding a failed part. In order for our policy change to be valid, Boeing must be able to catch as many errors as they do now if not more.

On the same token we seek to reduce the likely hood of Type I and II errors. If we widen our control limits we can reduce the number of Type I errors, while at the same time increase the likely hood of Type II errors, which is fine if Type I are much more expensive than Type II. Type I errors occur when an incorrect conclusion is reached that a special cause is present when in fact one does not exist and results in the cost of trying to find a nonexistent problem. Type II error occurs when special causes are present but are not signaled in the control chart because points fall within the control limits by chance. If we constrain our control limits or increase our sample size we can reduce our Type II errors, but at the same time, we increase the number of Type I errors. This is good because Type II errors are usually much more costly than Type I errors. However, as stated before, determining the cost of each error will help determine whether or not we should seek out Type I or Type II errors. (Evans & Lindsay, 2008)

CURRENT THROUGHPUT & COSTS

The system throughput must reach 122 stringers a day on average according to Randy Schimon. By analyzing the data we currently have, we have calculated that 44% of the time the system is in level 0 (average of parts and their overall states from the data given to us) and 56% of the time the parts are in level 1 or 1R. This means about 50% of the time, a part is going to be inspected. If we produce our average of 122 stringers in a 24 hour period, times 7 days a week we get 854 stringers, or 42,700 in a 50 week year. On average, you spend 4.7 hours inspecting parts ranging from 1.5 hours to 6 hours. This gives you an estimated 100,345 man hours of labor spent inspecting parts. By looking at this in a cost perspective, we can correctly judge whether or not this current method of inspection is practical compared to other methods that could be just as effective if not more.

NEW QUALITY CONTROL POLICY

Several things need to be measured in order to insure that the spars come out of the mills with all the qualities that make them the correct product. This process starts with the inputs into the system. These inputs include the program that is inputted into the mill, and the raw material put into the mill. The time is also an important variable to measure. The outputs that need to be measured are relatively simple. The spars need to be checked for width, height, and thickness throughout the entire length of the part as well as the overall length of the part. The total time it takes to create a spar, from beginning to end also needs to be measured along with the cost of the spar (including raw materials, labor and overhead). If all of these product attributes are managed correctly, it will ensure that the correct spars are coming out of the mill and through the inspection.

NECESSITY OF MEASUREMENTS

Beginning with our system inputs, it is important to have some measure of whether the correct program was put into the mill. This should be relatively simple because the program is just uploaded by the operator to the mill each time. Assuming that only one program for each part is in the system at any one time (Ex. when an update is created by engineering it replaces the old program), there really should be no mistakes at this point in the spar processing. It would also be important to check that the raw material going into the mill has the correct material properties as the desired material. If this was done in a manner that was quick and efficient you could easily avoid ever milling the wrong material and therefore wasting time in the process. Currently, this step is being completed by the suppliers of the material and they must be certified in order to supply parts to Boeing. If there is any problem with the wrong raw materials being delivered, Boeing would need to check all raw materials before milling or come up with a way to make certain that their current supplier or a new supplier only sent material that was within specification.

There are many inputs that go into the setup and milling process. These include mill setup, cleanliness, mill sharpness/maintenance, and the clamps that actually hold the part down. It would be very important to make sure the milling machine was setup and cleaned carefully before each use. Mill sharpness and maintenance would also need to be completed to insure that the machine could stay within its given tolerance and was not drifting due to a dull or worn piece. In addition, the clamps that hold the part would need to be functioning to avoid movement in the part during milling.

Outputs of the milling sequence are a very important measure even though they are not taken real time. The spars coming off of the mill need to match the shop drawings (for specific program) with a very low tolerance for error. This means several hundred measures could currently be taken from each inspected spar to ensure that they are of the correct thickness, length, width, and height. This is important not only because of the structural properties of the wings (ie Boeing cannot afford to install parts that cannot carry the loads for which they were designed) but also for installation (ie Boeing cannot just slightly oversize each part because installation off-size parts may be very difficult or impossible). Given these restrictions, Boeing needs to ensure correct thickness, length, width and height of the spars. Currently, thickness is not measured directly off of the mills because the thickness is changed many times throughout the rest of the spar creation process. We recommend that thickness actually be checked twice, once at the dimensional inspection process and once after the last process which changes thickness dimension. This would allow information to be gathered about what is causing thickness variation and possible part failures due to thickness. Thickness should then be measured first at the dimensional inspection point directly off of the mills because it is one of the quickest measures to take. This would allow inspectors to not waste up to six hours inspecting parts which will simply fail down the line due to the thickness inspect. If parts coming off the mill are passing at a very high rate at the dimensional inspection point but still failing at the thickness inspection point it may be acceptable to only inspect at the thickness inspection point. We believe this decision should be made by looking at concrete data instead of a guess.

Two other measures are necessary to insure the correct product is coming off of the mills. Time is a very important factor in the processing of the spars. The time it takes to mill and inspect a spar need to be recorded separately. The entire plant needs to be effective and fast enough to produce spars at a rate which will not hold up construction of planes without working overtime. This is the speed at which the plant needs to function. If the plant is pushing spars through at a rate higher than this, it really does Boeing as a whole no good. If this becomes the case, the spar plant could either be tasked with more work, could slow down and try to create higher quality spars, spend time and resources to improve their processes or possibly lay people off or cut hours in order to save money. It is also unacceptable for the spar plant to be the bottleneck for the actual plane assembly. If the plane assembly is being held up by the spar plant than the spar plant is delaying the sale and completion of a multimillion dollar product. For this reason it is important to measure the time it takes for spars to be milled (including setup, takedown, and mill time) and inspected in the plant. This will help create a benchmarking system the plant can look to in order to make decisions (for example, if engineering sends down a process improvement which saves 7 minutes of milling time but causes 3% more parts to fail inspection. We will essentially be decreasing the overall effectiveness of the inspection). The other measure needed to be tracked is the cost to mill and inspect each spar. If the spars are not being created cost effectively, than changes need to be implemented so the Frederickson plant isn’t costing the company money.

MEASUREMENT SYSTEMS

Currently there are no known issues with the raw material being delivered to Boeing so inspection of raw material is not necessary unless issues arise that would make Boeing suspect the raw material coming into the plant are out of specification.

There should also be some tracking system to make sure the part was put into the mill correctly, the mill was cleaned and maintained, and the clamps were effective at holding the part in place. A lot of these measurements will be done by the mill operator. They will oversee setup, cleaning and verifying maintenance is up to date. The clamps, however, should have a pressure gauge verifying a constant pressure was applied to the part and the pressure was sufficient to keep the part from moving during milling. The movement of the part laterally and horizontally needs to be measured with some sort of sensor which monitors the degree of part movement during the milling process, either from mechanical movement or from thermal expansion caused by the heating of the part during the milling process. Movement of the part during milling could be a cause of variation in the mill. If the part movement and expansion was monitored during the milling process, it would be possible to segregate variation in measurements due to part movement from the actual mill variation.

Some comprehensive system needs to be developed for recording each of the measures discussed above to make them traceable and readable for managers and quality control analysts. This should start at the input of the program into the mill. If the mill operator inputs the program into the mill, there should be a Quality Control file opened which matches the part. This file will remain on a server that can be accessed by several different people at the same time. The mill operator would need to input the part number, layout number, the file used to mill the part, and the mill used to mill the part. This would verify that the right program was being used to mill the part. In this program, the mill operators would document their mill setup and any changes they make to the milling program. Mill operators would also have no ability to change the NC program but would be authorized to change milling parameters after approval was given by engineering or another manager. All changes to the mill parameters would be documented so that both positive and negative changes could be tracked and used by other mill operators.

The current system for inspecting the dimensions of the spars is not effective at telling Boeing why parts are failing at the inspection point. The current system utilized is a system using tape measures and calipers, and paper tracking only if the part passed or failed the inspection. This gives Boeing little information about why and where the part failed. More importantly, upon a part failure, they do not need to inspect the next part coming off the mill because it will most likely be a different part number. This means that if the mill was causing the problem, the next part will fail also because the current process does not give Boeing any information about the mill. We believe that significant changes need to be made to this step of the measuring process. Currently, 3% of Boeing’s spars are out of specification. Given their current workload, approximately 600 parts per year are out of specification and missed using the current inspection process because each part is not measured. This is due to not all parts being measured. If each part isn’t measured, some parts will inevitably slip through the system and ship from the plant when they are an out of specification part. It is also shown through time series analysis that parts fail more often one after another. This can be seen in the Appendix under Figure A. Using Boeings current system, over twenty parts could come through each mill without one part being inspected. If Boeing is inspecting at Level One, they are inspecting a random sample of every five parts from each part number. Given that Boeing has several different part numbers; it is possible for a long string of parts come off one mill without any being inspected. Boeing needs to break out the mill inspection from the NC program inspection to effectively produce correct spars.

Because mill setup is completed manually, each part is essentially its own population of measurements. If mill setup could be automated to a high degree of accuracy between different operators, it could be possible to make the parts all part of the same population. Currently however, the parts coming of the mill all form a unique population but share a common population relative to the distance from the desired measurement. If we record the delta (specification – actual dimension) we can compare every part as a common population.

Ideally, real time measurements would be made on the spars before they even left the milling deck. Laser measurement systems are available that essentially ride along with the milling device and recorded real time measurements of what was actually milled. These measurements could be then uploaded into a cad file and a variation could be compared between the actual measurement and the desired measurement. These could then be tracked real time and compared to acceptable dimensions. If the mill was milling a part wrong the process could be stopped and the source of the variation could be found and fixed before more time and material was wasted on a part that would not pass inspection or be within specifications. We believe the technology is available for this type of measuring system and could be adapted to coordinate with the current milling equipment. This system could greatly reduce the need for dimensional inspection after milling is complete. A few spars would still need to be inspected to verify the accuracy of the automated system and mill.

There are Coordinate Measurement Machines that are available that could measure each spar with far greater accuracy and speed than an inspector with calipers. With some engineering, these systems could be integrated effectively into Boeings inspection process. Although we believe that the development of this technology would be worth the investment, an alternate system of measuring that modifies the current system needs to be put in place to improve and accelerate the inspection process until sufficient capital can be found to put in new technology.

The first change that we recommend is to measure the thickness at the dimensional inspection and again at the current thickness inspection point. This would add information about when the dimensional properties of the spars are being negatively affected and also save time by not further processing spars that are not of the correct dimension. The current method of inspection for thickness is acceptable using the ultrasound system. The current ultrasounds system utilizes a computer system and can measure the thickness of the aluminum much faster and with more accuracy than calipers. It is also very difficult to tamper with or mess up measurements due to the way the ultrasound is calibrated and the computer deciding if the part is within specifications. Laser systems could also be developed for Boeing’s use that could easily inspect thickness of each spar along with the other measurements.

Two of the other dimensions are currently inspected using calipers and have just started being recorded using pushbutton calipers. If calipers are continued to be used, these measurements must be recorded to the same computer file that the mill operator creates, as Boeing started doing between September 2008 and December 2008. The computer file should automatically take the engineering specification and the actual measurement and calculate a deviation from desired. The benefit of this is that the variation can be seen over time individual to each mill. This computer system could be automatically programmed to output control charts that have to do with variation for each mill as they mill each part and these changes could be automatically tracked over time. Once the parts are tracked, it would be possible to create control limits that would be set so that little or no scrap is created. If the mill is producing parts that are approaching the control limits (which would be stricter than tolerance limits), the problem with the mill could be addressed and solved before it creates parts out of tolerance. Boeing is essentially wasting time and data by not to recording these measurements and useing them as a tool for decision making.

In the current measuring system, the measuring of the length of the part serves a dual purpose. It determines that the part is of the correct length but also serves as a baseboard for where the other measurements are to be taken. As of right now, a tape measure is basically stretched across the part to check the length and then the same tape measure is used as basically a number line to locate the positions for the other dimensional measurements. This also compounds your error. If you only can measure to a certain accuracy with the calipers, the error introduced is compounded by only being able to be as accurate as the tape measure that is telling you where to measure. If a jig system was used instead it may improve the accuracy and repeatability of all the measurements. If each part was put in place and then butted up against a certain stopping point, you could have all the location measurements for each part marked on the table. It is our understanding that each part has a “pin point” on the inboard side of the part. This point can be used by the jig as a point of reference for all the other measurements. If a random sampling system was used, the jig could automatically light up at the certain points for measure saving the measurer the time of trying to find each point. This system would be simple and less prone to error. It would also eliminate error due to the movement in the tape measure or part during the inspection process.

MEASUREMENTS LOCATIONS

Due to the current system of recording, Boeing has very little information about where the problems are on each of their parts. This system is self fulfilling because if you have no data about where the parts historically fail, then you are left with little choice about where to measure on each part and therefore must measure all points on every part. It is recommended that Boeing continue the practice of measuring an extensive series of places on each part while recording these measurements and their locations. This will allow Boeing to build a database of problem places on parts, and see trends in part failure. This data may very well show that it is only necessary to measure a few key areas and at each end to check parts for dimensional accuracy, which would speed up the inspection process tremendously. It would be very probable that when setup and clamps are done correctly only the first and last points need to be measured. If those measurements are correct, the entire part will probably be correct. With the lack of data currently available about part failures, it is impossible to make these kinds of associations. Building a database of measurements will allow Boeing to make these decisions statistically instead of assuming unknown scenarios. There is also no data about what causes part failures (ie milling problems, chatter, bed problems, clamp problems, mill is dull…). Boeing has simply recorded that a part failed or passed. Without this data being gathered, it is hard to see how to improve the milling process. Recording the most probably reason for failure would allow Boeing to make assertions about how to improve the milling process. Our inspection process that we suggest to Boeing is discussed later in detail, but it involves breaking the stringer down into 10 foot sections and measuring randomly within those sections. This alternative allows a random sample that will likely pick up any variation caused by a clamp or holding failure along the mill. It also allows Boeing to view the milling process all throughout the stringer, which allows monitoring of multiple common cause variables such as the mill bit getting dull over time. The main focus of this measuring technique and locations is rational subgroup sampling, which is discussed more later, but essentially allows for any assignable cause to be discovered if presented because the chance of observing differences between samples is high, but the chance of observing differences within a sample is low, meaning that each sample is a true representative of the mill process at that time. All of those samples together present what the mill was doing over the entire stringer.

TIME OF INSPECTION

The inspection process is sufficient at the current location. A problem that does arise is not all parts are currently inspected and measured immediately when they come off the mill. This means that the mill could produce several more bad or failed parts because the parts were not inspected immediately and nobody knows the mill is off. This is one specific reason we suggest barring the possibility of real time verifications of measurements while milling is in progress. Raw material undergoes the milling process to cut it into the proper shape. Before it continues on with the production process it needs to be inspected to determine if it has the proper dimensions. Therefore the dimensional inspection process should be located after the milling but before all of the subsequent production actions. Its current position minimizes the possibility of traveled work and should reduce costs.

INSPECTORS

The measure should be performed by an individual other than the person machining the part unless they are measuring real time, then mill operator should do it, the reason not doing it now is not trust (should have an environment where it would not be an issue to have the miller measure), but he is busy milling the next parts while the others measure. The mill operator should be kept busy using his expensive skills instead of using a caliper. The purpose here is to create a separation of duties where the person inspecting the part has no incentive to falsely pass or fail for the part. The inspection is done by hand with electronic calipers. Inspectors will take measurements at various places on the part determined randomly by computer (this is discussed in further detail later in the paper). In the future, we recommend automating this process during milling in order to increase throughput by reducing cycle time and the number of parts milled incorrectly. Problems could be identified before the part became scrap because the measures would be compared against control limits instead of tolerance limits or specifications.

SAMPLE SIZE

Currently, the inspection procedure is divided into 3 levels. At Level 0, 30 consecutive parts are inspected individually for adherence to dimensional tolerances. After 30 successful inspections the process moves to Level 1 where one random sample is selected from a string of five parts of a given part. Any failures result in the level being changed to Level 1R, where a sample of ten consecutive parts is inspected. Any failures at the 1R level will result in the process being downgraded to Level 0. Any changes in the NC program will also create a mandatory Level 0. The current process tests to see if the program is correct by testing 30 parts every time a new program is created. The sampling plan will then test the mills variation in a non systematic way. With the current sampling plan, over twenty parts could pass through with no parts being inspected. Because the sampling plan only deals with certain parts, the mills variation is not systematically tested. This is a large issue because once the program is verified and is correct, the variation of the mills is what Boeing needs to test.

We feel this sampling method is not necessary. Essentially, two factors are being tested at this inspection station, the robustness of the NC program and the machine itself. The machine is tested for adherence to tolerances from the continual flow of parts through the milling process. The Level 0 inspection requirements for individual parts, in effect, test the NC program. When a new change to the NC program is implemented, a test sample could be taken to judge the program’s adherence to shop drawings generated by that program. We recommend completely inspecting three parts in order to statistically minimize the possibility of a type II error. This would test the communication between the mill and the program. If something was not milled correctly it would be detected before another part was milled. Engineering would be responsible for verifying that their program created the correct part, the dimensional inspection would verify that the mill was reading the program correctly. After verifying that the mill was reading the program correctly, all further inspections would simply test that the mill was within control. This inspection method is known as narrow limit gauging and due to the nature of the inspection process; the possibility exists of passing many failed parts. The samples we analyzed produced a 3% failure rate out of all inspected parts. This equates to 66 failed parts out of 3,000 inspected. However, 7,000 parts were passed through the mill, meaning over 100 potentially defective parts were skipped.

To rectify the high probability of traveling defective parts, we recommend a change in the inspection process. Every part milled should inspected with a sample size of 5 measurements, these measurements fall within a random 8 inch area selected out of 10 foot sections (further detail of inspection frequency is discussed in the following section). This provides four key benefits:

1. Total inspection time per part is reduced

2. Every milled part is inspected allowing the user to detect changes in the mill such as mean drifts.

3. The probability of traveling a defective part is reduced.

4. Accurate point in time milling process evaluation is viewed, with very minimal special cause probability occurrence within sample. (i.e. true indication of what milling process is doing at that time)

MEASUREMENT FREQUENCY

Measures relating to the mill before the milling actually occurs include calibration, setup, and vacuum holding. Calibration of the mill should be done at each shift change, and each time there is a different part number being produced on that mill. Taking the measures at each shift change not only lets the operator know that everything related to the mill calibration is in specification, but it also allows a control method of viewing if the calibration has moved or changed during operation over the time of the shift. It also moves the responsibility of correct calibration to each operator when they start, instead of the day shift operator being responsible for the correct calibration during the night shift. Recording all the specs on the results from calibration or any change made to adjust can put into a control chart. This would allow the operator and supervisors to see any movement and trends related to the calibration as the mill is used over the course of a shift. Any changes or adjustments in setup done by the operators or setup employees should be documented as well, so that the next operator or setup employees know that changes to setup for a specific part have been made. You do not need to calibrate the mill every time a part in placed on the mill, only at shift changes. If the time and cost of calibration is very small, then it is suggested that the calibrations be checked every time a new part is placed in the mill, especially if common cause variation is found and can be associated to mill setup specs or calibration.

Ideally, the actual setup and clamp holding for each stringer should be monitored continuously real time. Every time a new stringer is being milled, the setup and clamp holding of that piece should be measured to ensure it is setup and positioned correctly. Then the same measurements should be taken continuously during the milling process to ensure no movement has taken place. The clamps that hold the stringer down will be continuously monitored/measured by the suction pressure for each one; if pressure begins to drop then notification will be made to the operator via alarm. Continuous real time measurements will also take place that tells whether the stringer has moved in/out towards the mill and linearly up/down the mill, ensuring that that the setup of the piece has not moved. These could all be monitored with laser measuring devices attached to the mill bed or the mill. The real time data during stringer milling would be entered into control charts and allows any movement and trends to be seen while milling, which would allow operators to shut down the mill if the cutting begins to reach the control limits, and therefore preventing a possible failure. Alarms could also be set to warn mill operators when a piece has moved, or is reaching near control limits. A temperature sensor attached to the mill or mill head could read real time temperature of the aluminum as its being milled. This would allow control of variation caused by the expansion or contraction of the material. The temperature results would be documented as well so that they could be compared to the NC Program or operation setup. For instance if the data shows an alarming increase in material temperature during a specific milling point, the linear speed of the mill could be reduced or the amount of material cut could be reduced. Doing these measurements also helps rule out or associate mill calibration, setup, and vacuum holding as a possible cause of failure, that’s only if you record the data into control charts so trends in setup specs and recordings can be followed and compared to the control charts of the finished mill stringers (i.e. regular mill inspection results).

We would like to see real time measurements related to thickness, width/height, and linear features taken real time on the mill during the actual milling process by an automated measurement device. This would again allow for recordable data that can be put into control charts. Having control charts would give Boeing control limits instead of specification or tolerance limits and the advantage to that is any possible failure product or any movement towards out of control can be noticed before the piece is completely milled or milled into a failure.

Manual inspection of the stringers after they have been milled will occur for every stringer, although there will be two separate inspection processes based on a NC Program change and a regular inspection process.

After an NC Program change, the first three stringers produced from that program will be fully inspected with Boeings current inspection process. If all three stringers pass after the full inspection the NC Program is correct and regular inspection can begin. The regular inspection process (discussed in detail later) will ensure that the mill process is in control, only if Boeing monitors the control charts. Therefore, if the mill is in control, the three stringers will only pass inspection if the NC Program is correct because the NC Program tells the mill exactly how to cut. The reason for inspecting three stringers is because it removes the possibility of misjudgment based on various scenarios, which ensures that only correct NC Programs make it to the regular inspection process. For instance, one might be quick to assume that if the first of the three stringers fail, then the NC Program must be incorrect, but this may not be true because there is a minuet possibility that an uncontrollable special cause failure happened to occur for the first stringer produced from that new NC Program. Therefore, to not misjudge the program the second stringer must be inspected, if that stringer passes one can conclude NC Program is correct and a third stringer will be fully inspected to confirm this.

A second and very slight probability scenario exists if the NC Program is incorrect but the mill just happens to have a slight special cause variation, which offsets the NC Program’s incorrectness, and the result is a stringer that passes inspection. By having three inspected stringers we can make the correct conclusion on the NC Program because the probability of that special cause variation happening again is next to none, so the second stringer would fail inspection, and a third stringer would fail as well, which confirms the NC Program is incorrect.

If an NC Program is determined to be incorrect, change will be made and another NC Program will be implemented. At this point the same inspection process repeats again, as it does anytime an NC Program change occurs.

It is also very important that Boeing has a set of standard procedures for mill operators on the mill. To increase the integrity of the NC Programs, the mill operators should not have authority or be allowed to make any changes or adjustments to the NC Programs. Not only does this ensure the integrity of the program after it has been released by engineering, but it ensures that only the NC Program created by engineering is inspected for correctness, and if the NC Program passes with 3 correct inspections, then it ensures that NC Program won’t change later on during milling (without a specific purpose change done by engineering).

Regular inspection occurs when the NC Program is correct and we are now controlling the process of the mill by inspection. This inspection requires every stringer to be inspected but it will not be a full inspection as Boeing currently does. Every stringer will be placed on a jig; the jig will have linear measurements on it by foot and inches. Each stringer will be split into sections based on 10 foot increments. The first 8 inches and the last 8 inches of each stringer will always be inspected with 5 measurements at each portion. From all the sections a random sample will be taken. Where the random sample is taken within the section will determined by computer, but it will always be an 8 inch portion where 5 measurements will be taken. Therefore, the number of random samples or measurements will be dependent on stringer length. Any stringer or spar shorter than 10 feet in length will require an inspection of the first and last 8 inches as well as a random sample that falls between the two ends. Not all stringer or lengths are divisible by perfect 10 foot sections, so the last section (if not 10 feet) will be whatever is left over that is less than 10 feet. The same random sample and sample of the last 8 inches applies for this end section as well. For picture reference and further understanding, appendix FIGURE B shows a simple example of a 777 lower stringer broken down into sections with samples.

The computer will know the length of each stringer and break it up into the sections; the computer can then tell the inspector exactly which inch increments along the jig to inspect the stringer at and what those readings on the caliper should be. Or another solution would be to have a laser or light device, light up the exact spot that must be measured so that there is no error caused by the inspector measuring in the wrong spot. The reading from the caliper will go directly into a computer and the delta will be calculated. Having random samples within the 10 foot sections allows for control charts to be made based on the delta measurements taken inside the random sample portion. These measurements within each sample provide data and control to how the mill process was doing at that specific time. Taking the random samples across different sections then allows Boeing to see and control what the mill process is doing over a period of time and determine the robustness of the milling process.

This random inspection process based on sections also allows Boeing to see which areas of the mill create a larger deltas and it also covers vacuum/clamp areas, which would allow Boeing to see any irregular deltas related to possible vacuum/clamp problems as well. Each stringer or spar is inspected at the beginning and end, with a random (less than 10 ft. stringer length) or multiple random (10ft. or greater stringer length) inspections throughout the linear length. The idea here is that if the mill was in spec at the beginning of the process, throughout the middle of the process, and the very end of the process, then we can assume the mill was within spec during the entire process and on areas that didn’t get randomly inspected. This technique creates rational subgroups, where the chance of observing differences between samples is high, but the chance of observing differences within a sample is low, meaning that each sample is a true representative of the mill process at that time. In other words, this sampling method minimizes the chance of variance within a sample, while allowing for variation between samples to be detected.

All of the delta data recorded from the regular inspection will go into control charts where trends and movement can be seen, as well as control limits. Any movement towards the control limits, and Boeing can stop the process and fix the problems, preventing them from going out of tolerance and creating a failure. Boeing will also be able to look data from different sections and look for trends. For instance, if they see that between inch 400 and inch 500 that there is generally more variation or movement on the control chart records, they can closely investigate the milling process between those distances.

The electronic calipers used during the regular mill inspection should be checked at the beginning of every shift where they will be calibrated, adjusted, and zeroed out based on the manufacturer recommendations of the calipers.

Following these two inspection procedures discussed above provides opportunity to make sure the NC Program is correct, then go on to inspection of the mill process. The result of doing so is creating control charts that can be followed and prevent any failure from common cause variation occurring. The trends from inspection can also be compared to any trends with mill calibration, setup, or vacuum holding, which could simplify their search and investigation for any common cause variation that might occur.

Any failure that does happen following this inspection sampling should be the result of a special cause variation, which Boeing would fix if the special cause is assignable, obvious, or can be found. Special cause variations creating a failure can include inspector, inspection tool breakage, instant mill failure, mill bit breakage, untypical raw material flaw, power surge, or just a freak act of God/Nature.

QUALITY CONTROL RESPONSE TO MEASUREMENTS

The operators should follow the control charts, check out trends, shut down before data reaches out of control limits and explore the variation. Shut down for any special cause as well, if the result of the special cause failure is not an obvious immediate fix, then explore to find the cause. Compare the control charts together from different measures and look for any patterns with each other to help determine specific areas where the failure problem is coming from. If everything in the process before and during the milling looks to be in control, one might check the inspector. See if the inspector is carelessly inspecting the parts or not following inspections procedures. Or check the inspection tools make sure they are calibrated and create accurate measurement results.

DATA DISPLAYS

The data should be displayed in control charts via scatter plot style so trends and movement can be seen. Ideally these measures/control charts should be displayed real time as well, which will allow the operators and employees to see any variation as production is moving. This also prevents them creating a common cause variation failure by watching the trends and not letting any data go beyond the control limits. This helps them save money when failures (special cause) do occur because they know by the control charts that their processes are in control, and they don’t need to spend wasted time looking for common assignable causes. Instead they can investigate for the special cause if it’s not an obvious find.

DATA ACCESS

The shift supervisor and operators should have access to the displayed data. This would allow them to see how any adjustments affect the process or when they need to make adjustments based on the data on the control chart. Such as when the data gets close to the control limits or when there is sporadic variation moving in extremes around the mean. The supervisor and quality control can then make decisions on special cause and common cause variation. Then the correct actions can be made about what actions to take and where. The inspectors of the stringers after the mill should not have access to the control charts to prevent any biases as to their inspections on where they want to pass or fail a part (this is to prevent the Union inspection workers from creating job security by intentionally failing stringers). This can improve the robustness at the inspection stage as well, as the inspectors will probably be more consistent with their measurement techniques. All upper management and industrial engineers should have access to these control charts as well, especially when the charts show large variation or constant sporadic movement. If this were the case, then they might want to evaluate the process and search to develop new production techniques, procedures, ergonomics, or anything that would improve the process capability.

QUALITY CONTROL GOALS & CONCLUSION

The goal of our quality control process is to maximize the quality of the milled parts while minimizing inspection times and costs of error. Currently, the possibility of traveling defective parts is high and the possibility of moving between inspection levels is low. The goal is to improve throughput by instituting an inspection process that will minimize the possibility of traveling a defective part.

Also of importance, no formal rules exist for making modifications to the NC program. As previously stated, an NC program change will reduce the inspection level to 0 automatically. We suggest dividing the NC program changes into critical and non critical changes. Critical changes should be implemented immediately but non-critical changes should be worked into the production process in order to minimize the cost of additional inspections.

The result of the changes in the inspection process will create a measurement system that will allow the tracking of variance across machinery and people. The aim of the system will be to create a standardized logging system enabling documentation of mill setups and NC program changes which can assist in future operations. Ultimately, with improved documentation and a more robust inspection process, the goal will be to reduce labor and inspection costs.

APPENDECIES

FIGURE A

[pic]

FIGURE B

FIGURE C

WORKS CITED

Evans, J. R., & Lindsay, W. M. (2008). Managing for Quality and Performance Excellence. Mason, OH: Thomson Higher Education.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download