Measuring and controlling your metalworking processes is the best route to high quality production.

By Jack Teegarden
Federal Products Co.
Providence, Rhode Island

Many of the basic notions about quality in metalworking have undergone significant change in recent years. Where it once was enough to document part quality in order to satisfy demanding customers, ISO 9000 and the vendor certification schemes of various OEMs now make it necessary to document the quality process itself. Statistical methodologies have largely superseded inspection as the determinant of whether a lot passes or fails. Dimensional accuracy—once virtually synonymous with quality in metalworking—is no longer the only critical criterion. As tolerances in some industries enter the sub-micron level, design engineers are finding it necessary to write—and manufacturers to meet—detailed specifications for part geometry and surface finish.

Perhaps the most important shift involves the transfer of many responsibilities away from quality specialists, metrology labs and final inspection departments, and toward the machine operators on the shop floor. Quality is no longer a question of weeding out bad parts after the fact, but one of ensuring bad parts aren't produced in the first place. Quality is a major component of what the customer pays for. Thus, from the machine shop's point of view, it's a value-added activity, not unlike milling and turning. And the only way to provide quality cost effectively is to make sure it's produced simultaneously with the workpiece as an inherent part of the manufacturing process.

That said, it must be acknowledged that this idealized state of affairs has been attained by relatively few companies. In spite of all that's been written in trade journals, presented at seminars and expounded by consultants, most machine shops still operate according to quality paradigms that have been in force for fifty or sixty years, or at least ever since the basics of SPC began to be widely adopted.

Making the shift—converting quality from a cost-eater to a value-added service—is not an easy task. Above all, it requires a basic change in management philosophy about the purpose of quality, and the nature of the service or product being sold. Further, it requires unusual commitment to impart these changes to the workforce, and to make the necessary improvements to manufacturing systems.

When a process is "in control," with only random or chance causes of variation remaining, distribution appears as a bell curve.

Statistical process control is another essential link in the quality chain, but many who practice SPC assume incorrectly that it can make up for deficiencies elsewhere in the process. In order to provide optimum benefits, SPC requires good data to work with and a process whose capabilities are known and subject to accurate control. The "five-point plan" outlined here is designed to ensure that these conditions exist, making SPC a viable path to continuous quality improvement.

Step 1— Machine Tool Characterization

The need for machine tool characterization (MTC) is based on the idea that it's difficult, if not impossible, to get anywhere unless you know where you are. MTC is an essential first step that establishes a baseline for CNC machining accuracy. By measuring the machine's ability to position itself accurately, it is possible to take steps toward improvement, to know when those steps are successful, and to know when the practical limits of performance improvement have been reached.

Three-axis machine tools have 21 "degrees of freedom" which essentially means that they have 21 potential sources of positioning error. Each axis is subject to pitch, yaw and roll. Each contains some degree of error in linear positioning. Each has errors of straightness at two right angles to the direction of travel. And each axis may be out of square with the other two. In addition, the spindle may contain radial, linear and angular errors.

Without MTC up front, machinists can only treat symptoms of error as they appear in the workpiece. According to the old quality paradigm, machine tool errors are more or less expected and it is the machinist's job to produce good parts on machines in spite of them. This is inherently inefficient. As long as the underlying causes of error remain undetected and uncorrected, each setup will involve production downtime while the machinist tweaks the fixturing or the CNC code.

Four instruments are required for complete and efficient characterization of high precision machine tools:

The telescoping ball bar provides a quick, efficient means of testing overall contouring accuracy. With a couple of exceptions, the ball bar does not indicate individual errors, but rather provides a "snapshot" of the effects of combined errors, although some systems do incorporate diagnostic software to assist in machine repair. Two specific irregularities that the ball bar reveals directly are lead screw errors and backlash.

Electronic levels are used to level and straighten the machine tool with respect to gravity, which will often correct a host of workpiece errors. Electronic levels are inexpensive, easy to use and capable of measuring two types of errors that lasers cannot: horizontal axis roll and vertical axis yaw.

A laser interferometer system is used to measure linear positioning accuracy, straightness, parallelism and squareness between the various machine rails. It provides the most comprehensive set of tests, generates results to very high levels of accuracy, and can be used to test machines regardless of length of travel.

A spindle analyzer detects radial, linear and angular errors in the spindle. The spindle is the largest source of heat in many machine tools, and the spindle analyzer can detect "growth" of the spindle while it's running. This instrument distinguishes between errors that occur synchronously (consistent to every revolution) and asynchronously (randomly). If desired, tests may run several hours to analyze long-term thermal effects.

A practical approach to MTC involves total characterization of every machine tool in the shop on an annual or semi-annual basis. Every machine should be characterized when it is installed and again every time it is moved or rebuilt. Total characterization requires a level, laser and spindle analyzer.

In addition, the ball bar should be used to run quick performance checks on a more frequent basis as a preventive measure. With tests lasting less than a half hour, the schedule should be determined by the stability of the machine and the needs of the applications. In more extreme cases, tests may be as frequent as once a week or prior to every new setup. Ball bar checks should also be run any time there is a crash (or a suspicion of one), or if problems suddenly show up in parts. Should the ball bar reveal problems, then fuller diagnosis may be called for using the other instruments.

Step 2—Optimize Process Capabilities

Optimization, or machine tool calibration, follows directly upon characterization. The data gathered from MTC can be used to enhance quality, and in some cases, with little additional effort or expense. Here are several valuable uses of MTC data:

• Work can be assigned to machines that are shown to be capable of handling the accuracy requirements. Low tolerance work can be assigned to machines that cannot handle precision work efficiently.
• MTC data reveals sectors of machine axis travel and certain spindle speeds where performance is better than others. Operators can improve quality simply by concentrating work in these "sweet spots."
• MTC data is a useful tool for predictive maintenance. This is crucial because it costs far more to fix a problem after it has resulted in poor part quality than in a "preemptive strike" during scheduled maintenance. By tracking positioning accuracy over time through MTC, maintenance can be scheduled before the machine starts producing bad parts, but not so early as to create unnecessary downtime.
• MTC serves as a valuable troubleshooting tool, guiding the user to specific repairs or adjustments. Machines that do not meet manufacturer's specifications can usually be brought up to spec through calibration, while machines that are performing well can often be made to perform far in excess of manufacturer's specs. After maintenance is performed, tests are re-run and compared to earlier results to confirm that calibration succeeded in enhancing performance.

Every machine tool possesses a host of potential positioning errors. Shown here are all the possible errors associated with a single axis of motion.

In spite of the substantial processing benefits it confers, MTC has yet to be widely implemented outside of a few high-end industries that regularly need to push accuracy levels—most notably, aerospace and medical equipment. Most machine shops, it seems, are so eager to make parts that they won't take the time to make them better. The situation is analogous to the early days of SPC when shop owners said to themselves, "That's nice, but I don't need it yet." Thus far, only the most progressive and aggressive companies, and those that are truly committed to continuous quality improvement, are using MTC. All others are perhaps postponing the inevitable.

For those who have adopted MTC, the next phase is to establish a correlation between the results of static MTC tests and the part as produced. Laser and ball bar tests both occur with the machine in an unloaded state, and while these tests are extremely fruitful, errors detected by MTC do not always correlate perfectly with errors in parts. Proprietary efforts are under way at some companies to investigate this issue.

Step 3—Establish The Statistical Capabilities

The next step to achieve quality-inherent manufacturing is to determine the statistical capabilities of the process. This can be done only after the process is in control. "In control" is a state where only random or chance causes of variation are present. In other words, all assignable causes for variation have been found and eliminated. A production test run is performed, the parts are measured, and the results are graphed. When a process is in control, variation assumes the familiar bell curve distribution. If the test run does not produce this form, hidden sources of assignable error remain to be found through further application of MTC.

Machine or process capability is viewed within the context of the desired product spread or tolerance. As recently as the 1980s, it was usually acceptable to run a process where the + 3 sigma spread (plus or minus three standard deviations from the mean) was 75 percent of the product tolerance. Today, under the pressures of continuous quality improvement, 50 percent is more representative. This allows the process to increase somewhat in spread or drift from the mean, yet still remain within tolerances.

If the process spread is too wide, even when the process is in control, three options exist:

• Ignore the chart, let the chips fly, and accept a high scrap rate. Some companies that produce large quantities of low-cost commodities find this an acceptable way to do business, but it's the very antithesis of a quality-inherent approach to manufacturing.
• If the spread is only slightly excessive, a rigorous inspection protocol can be implemented to catch process trends early enough to keep the process in tolerance. This might require 100 percent inspection, which may be practical for small production runs, but inefficient for large ones.
• Use a different machine tool with the required capabilities for statistical control. This is the only approach that will help achieve the goal of quality-inherent manufacturing.

Step 4—Specify Inspection Mode, Equipment And Procedures

Of the many metal removing processes, only circumference grinding lends itself to real-time process control. All others require post-process gaging, which should be followed as soon as possible by the application of statistical process control.

A worker prepares to test a machine tool spindle for radial and longitudinal errors that could produce irregularities in machined parts. Characterization and calibration of machine tools is a key, but often overlooked, step to continuous quality improvement.

Whether it is performed in-process or post-process, inspection requires process engineering similar to that which applies to manufacturing. Applications vary so widely that it is impossible to establish an inspection regime that combines accuracy, reliability and economy without considering the following variables:

• Number and type of features or characteristics to be inspected (diameter, height, thickness, location, squareness, parallelism, roundness, concentricity, and so on).
• Throughput requirement.
• Required level of accuracy.
• Full 100-percent inspection, or audit methods?
• Level of operator skill.
• Portability (of gage versus workpiece).
• Inspection environment (including cleanliness, temperature, vibration and other ambient conditions).
• Workpiece condition (cleanliness, temperature, geometric variability, surface roughness, presence of flash, and other such factors).
• Workpiece material (Easily scratched? Compressible?)
• Nature of the manufacturing process (different processes impose different types of error).
• Data output format required.
• Budgets for gage acquisition, maintenance and inspection.

The salient fact of the "Great Quality Shift" is not the transfer of gage hardware from an inspection department to the shop floor but, rather, the transfer of responsibility for quality assurance and its metamorphosis from a post-process activity to one that occurs pre-process and in-process. Machine operators must, therefore, be educated in the principles of inspection and quality assurance. They must understand the nuances of gaging, including the effects of contamination and temperature, and the importance of proper gaging and mastering technique. They must be able to distinguish between part error and inspection error. Whether or not automated SPC is in use, they must have a working knowledge of the subject in order to recognize process trends and take appropriate, timely action.

Reeducating a workforce and imposing additional work responsibilities is difficult but rewarding. When successful, it can reduce the inefficiency inherent in having one individual responsible for the quality of another person's work; it can save costs through the reduction of scrap and rework; and it can lead to continuous quality improvement. It provides the individual employee with the gratification of becoming more competent and knowledgeable. And, ultimately, it may be necessary for company survival.

Step 5—Confirm And Maintain Inspection Capabilities

The validity of the inspection process itself must be confirmed and maintained. At the very least, the condition of the gages must be monitored, and repairs made when needed. This is universally accepted by quality-conscious companies where programs of periodic gage calibration are well established.

With the growth of shop-floor inspection, however, it has become more important to confirm the entire inspection process. This is performed through gage capability studies, usually known as GR&R (gage repeatability and reproducibility).

In GR&R, multiple parts are inspected multiple times by multiple gage operators, then the results are reduced to a single numerical value representing the variability of the inspection process as a percentage of the total part tolerance. While many companies use GR&R for acceptance testing of new gages, few take the process to the next logical step and apply it to existing inspection processes. This is unfortunate, for a program of regular GR&R lends a high level of assurance to shop floor inspection and is useful to identify areas where procedures need improvement or special monitoring.

In order to obtain the maximum benefit from a GR&R program, users must be aware of its limitations. A part feature with inconsistent dimensions will produce inconsistent gage readings, for instance, unless the part is staged with absolute consistency from trial to trial. GR&R fails to take into account the effects of within-part variation, and does not address the need for consistent staging. Thus, many users fault the gage for inconsistent results when in fact the gage is providing an accurate indication that the manufacturing process is inconsistent.

Within-part variation takes two common forms: geometric variance and surface finish effects. A nominally round part, for example, may exhibit out-of-roundness when analyzed on a circular geometry gage. But if measured multiple times on a simple diameter gage, without taking care to repeat the gaging position each time, the within-part variation might be misinterpreted as poor gage repeatability. Likewise, the minute peaks and valleys that make up the surface of a machined part may influence some gage results. Users should be prepared to test for geometry and surface finish effects on dimensional consistency when GR&R indicates excessive inconsistency in the measurement process.

Application Of Statistical Methods To Process Control

An ongoing program of gage calibration is essential to ensure the accuracy of the inspection process. Here, a master ring is calibrated in an in-house metrology lab using a ring-and-disc comparator.

With inspection procedures established and confirmed, the machine shop can begin making parts and controlling the process using either of the basic approaches to SPC: pre-control or control charting. Fortunately, most companies have grown well acquainted with the practice of SPC, making discussion of its methods unnecessary here. But SPC cannot by itself assure good product quality. It requires a number of preconditions to be effective:

• The process must be controllable.
• The process must be stable.
• Data to assess the process must be readily available.
• Data to assess the process must be accurate.

The five-point program outlined here is designed to establish the necessary preconditions that can make quality-inherent manufacturing a reality. Such a program requires commitment, but will yield valuable payoffs. Workers' morale is likely to improve as it becomes clear they are truly in control of the quality of the company's products. Scrap and rework will decline, and overall conformance to specifications will improve. Machine tool utilization and productivity will increase, and the costs of quality will recede. Finally, the program will enhance marketing success as the company will be able to demonstrate that quality is inherent in the process and the product.

For more information from Federal Products call (401) 784-3100, visit their Web site at www.fedgage.com, or select the Product Info icon at right.

Serving the Metalworking Industries. Since 1928.
MMS Online and all contents are properties of Gardner Publications, Inc.
All Rights Reserved. Reprinted by Permission.

Email: info@buckleyowens.com