For high-value aerospace parts, the emphasis is on
By Jim Destefani
"You can't inspect quality into a product" is an old axiom of manufacturing that's been taken to heart at GKN Aerospace (St. Louis), where the emphasis is on knowing machine capabilities up front to limit production of any bad parts.
Essentially a large aerospace job shop, the plant manufactures machined and sheetmetal parts for the F-22, F-15, and F-18 fighters; T-45 trainer; the C-17 cargo plane; and other US military aircraft. GKN's 750,000 ft2 (70,000 m2) machine shop houses, among other equipment, 54 large three- and five-axis gantry machines with a total of 162 spindles and bed lengths to 120' (36.6 m).
The gantries formed the core of GKN's machining capabilities for aluminum and titanium alloy parts, but most of them were vintage 1970s machines and some had seen better days. Faced with the challenge of improving its machining capabilities to meet tolerances required for a new generation of aircraft such as the F-22, GKN engineers chose to rebuild/retrofit some of the gantries as well as bring in some new equipment.
To make the most of current machine capabilities and decide which machines to retrofit, managers instituted a machine capability testing program. Performed before any parts are machined, the capability studies have multiple benefits, according to quality specialist Ulysses Green. "The capability studies really determine the machine's potential accuracy without ever cutting a chip," he explains. "They allow us to determine what the capability of the machine is based on machine motion."
GKN's massive machine shop houses 54 three- and five-axis gantry machines with a total of 162 spindles. Many of the machines are now being rebuilt and retrofitted to improve their capabilities.
Armed with that knowledge, engineers can then match job tolerances to machines with more precision. "For years, any supervisor or seasoned machinist in any shop had an intuitive understanding of which machines could hold which tolerances," Green says, "but they didn't know why. These machine validation studies allow us to assign a number to that, so we can quantify what we're able to do and put jobs on the machines accordingly. And, if we know the machine's inherent accuracy, we can also make allowance for other variables, such as cutter runout." Depending on the part, tolerances can be as tight as ±0.004" (0.1 mm), he adds.
This portion of a spreadsheet shows how data from machine capability studies are reported. Although data have been deleted from empty cells, the form's color-coded format facilitates quick diagnosis of problem areas in the machine's work envelope. (Click on image to enlarge)
Finally, the studies help in root-cause analysis in the event of a problem, Green says. "You can cut a part and have a failure, but then you'd have a hard time trying to determine what caused the problem," he says. "Was it spindle runout? Was it the program? Was it a worn cutting tool? There could be a variety of causes. Using this method, we can eliminate the machine as a possible cause."
GKN's machine validation efforts revolve around use of a terahedral artifact placed at multiple locations on the machine bed. Quality personnel perform 84 measurements--one on each corner of the tetrahedron--in 21 locations on the machine bed in less than two hours, according to Green.
Called the Spatial Reference System, the artifact is supplied by a company called Metronom US (Ann Arbor, MI; see sidebar for more information). "The shape of the artifact and how we use it gives us squareness, straightness, scale, and all the measurements you'd get off a CMM, but a lot faster," Green says.
"So we're working with our maintenance team to footprint each machine, and we actually get a volumetric accuracy number for each machine. The number that we get out of our process is inaccuracy in X, Y, Z, A, B and everything else all rolled into one number," he says.
The actual checks combine procedures recommended in the ANSI B-89 standard for calibrating CMM machines and those contained in a corresponding VDI/VDE standard. Green says the configuration of the artifact dictates this approach.
"The B-89 standard calls for probing several points on spheres, while VDI/VDE specifies single measurements on step gages," he explains. "We're checking all the things outlined in the standards--squareness, straightness, scale, etc.--and we're taking both sphere and single-point measurements.
"Of course, CMM machines are three-axis machines," Green continues. "Most of our machines are five-axis. So we came up with our own method of checking the additional axes using laser measurement devices. We can thoroughly check a machine, all three spindles, in less than two hours, and we get a lot of meaningful information."
If Green's team--currently, four people are assigned to the validation project--finds a problem with a machine, it is turned over to GKN's machine maintenance department. "They determine the root cause and corrective action needed," Green says. "They're very good at doing that and documenting the fix. Then we can come back and check the machine again if necessary."
Data analysis and reporting also combine multiple approaches to give GKN engineers, managers, and other personnel a useful snapshot of machine capabilities. "We take all the data from our tests and bundle it together using multiple software packages," Green explains. "The most important are the Metronom software and a package called SpatialAnalyzer [3-D graphical metrology software from New River Kinematics, Williamsburg, VA]. The two packages work hand in hand, and we had to develop our own process to get the system to work."
Data are reported on spreadsheets with a format similar to that partially reproduced in this article. According to Green, the column headed "Spindle #/Location" indicates the spot on the machine bed where the measurement is taken; the numbers A-1-51, etc. indicate the corner of the tetrahedron being measured. "'Nominal' is the calibrated length of the bar on the artifact, and upper and lower limits vary from about ±0.004" (0.1 mm) to ±0.030" (0.76 mm)," he explains.
A total of 84 measurements are logged for the calibration procedure. Data are calculated in terms of both deviation from nominal and as a percentage of the tolerance band, then marked as an "X" in the appropriate box in the colored portion of the spreadsheet.
The color-coded spreadsheets are based on a statistical process control (SPC) approach originally developed for small lot sizes, which can hamper SPC efforts because jobs produce an insufficient amount of data to allow good estimates of process control limits, mean, and standard deviation. Called percent tolerance precontrol charting (PTPCC), the method was developed by S.K. Vermani of The Boeing Co.'s St. Louis supplier quality management group.
PTPCC is said to work well for high-capability processes, although process capability does not need to be known in advance. The procedure divides a standard precontrol chart into five regions--a central green region, two yellow regions on either side of the green region, and two red regions on the remaining sides of the yellow regions.
Centered at the nominal value of any dimension, the green region covers ±50% of the tolerance band for the dimension. Each portion of the yellow region is half as wide as the green region, so the two areas together cover the full width of the tolerance band. Red regions are therefore out of tolerance.
To use the system for process control, operators check five consecutive pieces and implement a chart if all five are in the green region. They then select two consecutive pieces from each production run. If both pieces are in the green region, or if one is in the green and another in the yellow, no action is needed.
If both pieces fall in the yellow region, the process is adjusted and the operator attempts to initiate a new chart. The same would apply if one or both pieces fell in the red region.
PTPCC works for short production runs and results in a visual picture of process capability, which appeals to Green. "The good thing is, it works for small lot sizes. And, it's very easy to look at the chart and see what's taking place," he says. "Everyone can see the tolerance bands, the location on the machine table, etc. The color coding makes it simple, too.
"You start talking about things like Cpk and everyone thinks or acts like they know what it's about, but they don't always know," he adds. "But it's very easy to follow when you do it this way, and in fact the results are the same, because a Cpk of 1.33 corresponds to 75% of the tolerance band."
Up-front quality makes sense, but customers still dictate inspection plans for finished parts. Rebuilt/retrofitted gantries include updated CNCs with support for spindle probes, and GKN makes use of this capability to assure in-process quality.
"We minimize error by introducing probes," Green says. "For example, we machine the first side of a part, and in that process we put some holes or other features in the part. We indicate those features using the probe, then we flip the part over and re-indicate the features for alignment."
Operators also gage certain part characteristics in process--for example, flange thickness or other key features--using dial calipers or micrometers. But the emphasis is on producing parts correctly the first time, and GKN hopes that by doing so it may someday be able to reduce post-process inspection requirements.
"Inspection instructions vary from part to part and customer to customer, but our focus is on verifying that what we're doing is correct--the NC program is correct, the machine is performing as designed, etc.," Green concludes. "If we're really able to look at our machines and get them all running with good capability, then we may be able to go to a sampling plan for parts as opposed to the current 100% inspection."
|Pyramid Reveals Machine Secrets
One of the keys to GKN's machine validation program is use of a tetrahedral artifact that allows fast checks of all axes on the plant's three and five-axis gantries and other machines.
Supplied by Metronom US (Ann Arbor, MI), the Spatial Reference System is designed to validate production and measuring machines, including CMMs, robots, and machine tools. The system's simplest configuration is a bar with one magnetically attached steel ball at each end.
The Metronom Spatial Reference System
Users can quickly combine multiple bars to build precise 3-D shapes. Metronom says the tetrahedral configuration, built up from six bars and four balls, is the most applicable shape for many 3-D applications because it allows six fast length comparisons while providing compensation data to the system being checked. Other possible 3-D structures include octahedrons and icosahedrons (shapes with eight and 20 equilateral triangles as sides, respectively). The company also points out that for three-axis machining center applications, a 2-D network of triangles often is the most appropriate configuration because of the machines' relatively short Z-axis travel.
The SRS is available with standard bar lengths of 500, 1000, 1500, and 2000 mm, and the company will produce custom lengths as well. According to Metronom, the SRS is more suitable for qualifying large-volume CMMs and gantries than step gages and other devices.
The steel balls can be replaced by probing targets, allowing use testing optical measurement instruments such as laser trackers or theodolites. Applications include correlation of CMM and CNC machines; determining appropriate calibration cycles for CMMs and machine tools, and optimizing CMM performance; providing rapid checks of articulating arm portable CMMs, laser tracker, systems; and other devices; and referencing industrial robots. Metronom says the SRS can also serve as a tool to train new operators of CMMs or other measurement systems.
The artifact is used with Metronom's SRS-E evaluation software, which gives users NIST-traceable data on actual machine performance. The package uses Intelligent Optimization, Regulation, Adjustment (INORA) algorithms, which are said to provide an error-robust evaluation of measurement data.
This article was first published in the March 2005 edition of Manufacturing Engineering magazine.