Printer Friendly

SPC for zero defects.

SPC for zero defects

If you can prove your manufacturing process is in control, you'll make all good parts, and your customer will know it. You'll both save time on inspection and rework, and the quality-control system will pay for itself sooner than you think.

But can we really have zero defects? When the ZD flag was so popular a few years ago, a lot of the emphasis was on people. Unfortunately, the goal was impossible, because people are only human.

Today, good people are still essential for quality assurance, but the computer can provide most of the assurance in QA. The low-cost microprocessor can back up the human effort, and thus ZD is now a practical goal.

Much of the new look in QC is the result of statistical analysis of inspection information with instant feedback to the production line. For our discussion, statistical process control (SPC) in metalworking has two forms: (1) Real-time feedback of analyzed inspection data to directly control toolsetting while the tool is in the cut, and (2) fast analysis of inspection data to automatically reset the machine or tools before making the next cut, and before parts go out of tolerance. It sounds like adaptive control, but it's much more sophisticated.

Beyond feedback

Statistical systems discover trends in parts production and warn operator and automatic toolsetting systems that machines are heading toward making parts at upper or lower tolerance limits--or that the tolerance spread is too wide. Often, simple adjustments keep the machine safely within the tolerance bandwidth.

If the process is truly in control, it will not be necessary to use 100 percent inspection. Sampling an occasional part at random will suffice. Of course, a broken tool or machine malfunction will automatically shut down the process and flag any bad parts.

The statistical element puts more thinking or reasoning in the computer control. It knows that an occasional bad part (caught and flagged) is acceptable, but three in a row is grounds for shutdown! It gets to the bottom line of interpreting the various control charts familiar to QC engineers.

Air Gage Co, Livonia, MI, calls it Computer-Aided Gaging (CAG). "It's basically checking for a bad process instead of bad parts,' says William Duey, VP of the firm. "We build our systems from standard modular components, hardened for survival on the factory floor. The systems employ floppy disks or Winchester hard drives, and once the hardware is in place, software changes are relatively easy for upgrading the system or adapting it to new workpieces.'

Transmission-case analysis

Air Gage Co recently designed and built several CAG systems for the automotive industry. For example, Figure 1 shows a computerized gaging system employing 88 probes in 18 spindles to inspect 83 close-tolerance characteristics of three models of an automatic-transmission case for front-wheel-drive cars. The case, Figure 2, must be checked for multiple relationships, sizes, and locations of details in all models.

The gage system is considered "off line' because it does its work after the part is completed, even though it is physically on line and designed to work fast enough to statistically control all close-tolerance machining operations preceding the gage station. As in the case of most SPC operations, it uses previous inspection data and a gage computer to provide statistical process control of machine tools in the production line.

The system consists of five mjaor components:

Gage fixture and probes.

Computer for storing, manipulating, and communicating data.

Modicon 484 programmable controller to control all gage functions.

Complete set of masters consisting of hand-held minimum-and-maximum masters to set gain, and a mean calibrated master to work with the computer to offset and zero the data.

A 200-gal hydraulic system to actuate various movements of the gage. Also included in the system, but supplied by the end user, is a Lamb Cyclobot robot for loading and unloading the gage. It takes five consecutive cases off the production line for each sample, stores them, and loads them one at a time into the gage.

The main gage is housed between two granite blocks on the top and bottom to minimize any gage movement caused by changes in plant temperature. Also, the granite blocks are isolated from the gage base and the floor by isolation pads to enhance gage stability. The probe, Figure 3, uses a half-bridge LVDT. The designers feel that a single-coil model is more accurate than a double-coil type, and, of course, it's cheaper. A strain gage can also serve in some setups.

Parts are automatically loaded onto a pallet over two locating pins and hydraulically clamped on three locating pads. Air sensors assure proper part location to avoid gage damage.

After part clamping, the pallet moves under a bar-code scanner that reads and stores an individual ID for each transmission case. Information gathered by the gage is logged against a bar code for future use by service and QC departments.

The pallet and workpiece are then raised into the gage station and located against four pads and over two locating pins in preparation for gage-spindle actuation and eventual reading and storing of data. Analog signals obtained from the gage are transmitted to a Dactron 1 standard computer interface, Figure 4, for conversion to digital signals and transmittal to the computer. Once the workpiece is inspected, the pallet will return the part to the load station for unclamping and automatic unloading. One complete cycle of the gage is about 90 sec.

After the parts are gaged, data will be stored and analyzed by the computer to determine whether the process is in control. The computer calculates and stores the accumulated data from each five-piece sample and combines the most recent data with that previously acquired to determine tolerance trends. The computer does all this in real time. It can provide the charts shown in our title photo, but human analysis of the charts would not be fast enough in most cases to prevent production of bad parts. It's the instant feedback of this information that makes SPC viable. The printouts are merely for the record and the customer.

The computer system consists of a standard Hewlett-Packard Model 9920 (a 16-bit model with a graphics printer), a 12 CRT, and HP 3497 data logger, and an HP Winchester 15-megabyte drive in a sealed enclosure to withstand factory environments. A remote, membrane-sealed, numeric Keypad aids manual operator interface.

The automatic gaging system can master itself by using minimum and maximum gage masters, all directed by the computer with its related software. A jib crane on the side of the main gage helps load and unload masters and various gage components. The entire gaging system is housed in an area of 10 ft wide 24 ft long.

Air Gage Co also built an 11-station in-line automatic gaging system for inspecting crank- and wrist-pin bores of connecting rods at a rate of 1800 pcs/hr. Again, it inspects the conn rods immediately after finish boring, so manufacturing can make necessary maching corrections before parts reach an out-of-tolerance condition.

The rods feed into the gaging system, crank end first, by a gravity load chute. A part escapement mechanism in Station 1 releases the parts one at a time into a chain-driven synchronized transfer. This shuttles the part into a gaging fixture, Station 5, as shown in Figure 5. The part is located by spring loading its thrust face against three pads on the face of the crank-bore probe spindle. Once gaged, the resultant signal is processed for acceptance by a 16-bit microprocessor.

The conn-rod system isn't asking for perfection from production. It sorts any defective rods into several bins, based on measured defects or wrong sizes!

Don Regan, process engineer at Ford and the man responsible for successful utilization of the Air Gage Co equipment, notes that the mathematical concepts are nothing new. The X-bar and R charts have been around for a long time. In fact, some of the production ideas we are "pioneering' now are concepts we gave the Japanese in the early '50s and then dropped ourselves.

Regan says that, although the charts are old, the approach is new. The computerized gage does what a man can't do: It records low and high limits and forecasts trends. "A human is likely to overreact if he finds a poor measurement. He'll adjust the tools too soon. But a process chart does not overreact. For example, if the computer finds that parts are generally undersize, it suggests that the operator use manual gages and check for problems in more detail.'

The CMM approach

So far, we've looked at special machines loaded by a dedicated robot. Statistical analysis for SPC can be based on simple inputs from electronic snap gages, height gages, or any other source, as long as measurements are taken quickly when the parts come off the machine tool. But the coordinate inspection machine can serve in and SPC setup, too. You don't think of thorough inspection as fast work, but Digital Electronic Automation (DEA), Livonia, MI, has several measuring systems designed for in-line applications.

The firm offers motorized bridge and gantry measuring machines that use computer control to move the probe over the workpiece. For many setups, this can bring the measuring process up to speed so it can sample batch parts for SPC. For faster inspection of fewer dimensions, DEA's Bravo measuring robots provide 100 percent on-line inspection in medium-variety series and batch production. They can cycle at 120 measurements/min.

Once again, the system monitors the production line, preventing defective parts during manufacture, rather than simply rejecting them after the process is over. The concept features a modular open structure that can be integrated into to manufacturing line and configured in multiarm cells that combine the flexibility and programmability of the measuring machine with the speed and ease of use of conventional inspection gages.

Clark Hibbard of DEA says a lot of people talk about statistical evaluation leading to process control, but the trick is how to apply it to an automated system. He says, "Catching the drift is easier with a lathe, because it usually has only two axes with which to contend. It's harder with a machining center. You have to ask, is the hole in the wrong place, or is the reference wrong?

"The problems are underestimated when people go into automated gaging. Statistical-control algorithms are different for each application. What percentage of allowable tolerance will sound an alarm? How will you calculate the data? Are you auditing the process or the part? Your customer will establish the protocol. One user might want the system to stop everything if a bad part is found, another might simply want the system to identify that part and its bad feature.'

Other tricks of the measuring trade result from the fact that we are just beginning to learn how machine tools react to temperature changes and clamping stresses. The part might be a different size on the leeward side of the machine, and it might change size after removal from the fixture. Sometimes it is necessary to measure a part "off' to compensate for changes after the "fixture squeeze.' Or it might be necessary to let go of a workpiece and reclamp before finishing the part.

You can see that not all CMM operations are going to be fast enough for SPC, unless the data are entered on an audit basis. DEA engineers consider 0.0005 a close tolerance for a CMM, but the firm offers a small bridge machine with capability of 0.0001 if required. Hibbard says, "Our competitors claim similar tolerances, but you should ask them if they make their calibrations according to the B89 Committee standard. It's one thing to claim a machine is accurate when measuring parallel to the axis anywhere on the bottom, but yet another to guarantee accuracy anywhere in the volume, using a ball bar. This latter measurement checks backlash, hysteresis, and straightness of travel.'

Finally, some CMMs are designed only for light duty, others for true production work. Any measuring robot or CMM must give repeatable measurements, or the resulting statistical analysis will be no better than a 10-day weather forecast!

Monitoring for SPC

It's beginning to sound like the old computer story. Statistical analysis is find, but, beware of garbage in, gospel out! This is where people still count. Engineers must select proper equipment that will truly do the job; operators must make sure they input every detail; and management must make sure the system is really producing good parts at a profit.

From management's viewpoint, there is a need for the big picture. The Cross Co. Fraser, MI, offers equipment to project this image. Statistical production data and machine-diagnostic information are provided for transfer lines and medium- and high-volume transfer systems by a computer-based system called the Sentry V machine monitor. A machining system using this monitor appears in Figure 6.

The monitor is built for use on the shop floor. It has sealed keypads, protected screens, and a nonvolatile bubble memory that is not affected by power losses and that doesn't require air conditioning. The system provides a graphic plan view of the particular machine it's monitoring and therefore replaces the map lights normally used. It produces detailed operations charts and graphs on its 19 color CRT, and alphanumeric lists and statements on its 12 monochrome screen.

The monitor generates many documents, including a production report, production history, tooling report, tool-change counter list, fault history (last 500 faults), and fault histogram--a bar graph of 20 most-frequent faults and resulting downtime. Fault Search locates trouble spots and provides fault history, and a time-line bar graph indicates cycle time--the actual time of the last cycle compared with a standard.

The reports assist SPC programs, and the system monitor coordinates and processes data from other machine monitors for centralized management of production control.

Interface for SPC

For a successful SPC system, getting data to the computer is half the battle. From a as many as 96 measuring instruments, located up to one mile away, it's possible to transmit measured data to a host computer. The connection between measuring instruments and computer is provided by an MUX-40 interface system from Mitutoyo/MTI Corp, Paramus, NJ, see Figure 7. The system accepts data from Digi-Matic calipers, micrometers, indicators, linear gages, micrometer heads, and height gages. The data proceeds through a switch box and transmitter to an MUX-40 main unit equipped with either a GP-IB or RS-232C interface for connection to most computers.

Brown & Sharpe Mfg Co, North Kingstown, RI, has a practical idea for SPC. The PASS system (Plan for Analysis by Statistical Sampling) employs gages that enter workpiece measurements directly into a computer, without manual keyboarding, for immediate statistical analysis. Figure 8 shows the basic components that eliminate the time required for recording and calculating measured results. The equipment is designed for job shops as well as companies with established systems for statistical analysis. Where QC facilities are limited, PASS eliminates manual calculation and provides documentation to keep control of productivity and manufacturing costs. The gages are familiar hand surface-plate tools including Digit-Cal universal digital calipers and Digit-Mike II electronic digital micrometers. Transmission is by RS-232C format.

DataMyte Corp, Minnetonka, MN, offers similar equipment, Figure 9, with emphasis on the point that nobody wants to manually plot control charts, even though the charts should be maintained right at the process. The FAN system provides the charts without repetitive math or paperwork. For a copy of the Factory Area Network Buyer's Guide and the DataMyte Handbook, a 432-pg publication teaching basic SPC theory and methods of data collection, with case studies and details about SPC products.

Software and hardware

SPC software written for the IBM PC is available from Business Systems Design Inc, Menomonee Falls, WI. During development, the software was tested in a precision machining application. Inprocess and historical data can be used for process-capability studies, and results are available in a plain or graphics mode.

The software now provides batchmode data entry, but will be expanded to include use of electronic gages for direct input from machine to stand-alone workstation and for automatic compensation of the machine tool. The software is the first phase of a total-plant master system that will collect and store data from stand-alone workstations.

If you install your own SPC system, you will need a source of self-compensating tools. Samsomatic, Detroit, MI, offers fixed tool positioners, rotary tool positioners with amplifiers, and automatic tool and gage controllers as part of their line of tool-correction systems and components. Their systems can be designed into new machines, or retrofitted to older finish-turning or finish-borning machines. The firm says that retrofitting is practical, even if the machine is not designed to accept tool-correction equipment. They offer everything from tools, holders, and positioners to gages, amplifiers, and both pneumatic and electronic controllers. Microprocessor-based systems are also available. Systems will correct for edge wear of inserts on fixed or rotating tools, correct for thermal effects as a cold machine warms up, and cut frequent shutdowns for insert adjustment.

Auto tool correction can increase production rate, maintain accuracy, reduce trial runs, and allow automation where tool wear was a limitation.

Selling the concept

If you thought this discussion was academic, take a look at the specs coming down from your customers. They want more than quality parts; they want proof. Many small shops don't know whether they meet standards, because they don't have proper documentation. They rely on the boys in the shop to make it right.

But today, and certainly tomorrow, you will need proof of good QC systems. If you have automatic screw machines, you can kill two birds with one stone. Metal Seal & Products Inc, Willoughby, OH, is a large screw-machine shop making precision parts for automotive and other equally critical customers. They learned so much about statistical QC concepts, e.g., control charts, bell curves, and sigma, that they give seminars to other plants, including those of Alcoa and Parker Hannifin. They also can teach the small shop how to set up a quality program. For details, contact Plant Manager Russ Diemer, or Quality Control Manager Dave Saletrik, phone 216-946-8500.

The basic concern at Metal Seal is the answer to the question, "Is the equipment capable of maching to the tolerances needed?' If the answer is yes, you have little to do but set the tools properly and check an occasional random part. If the answer is no, you better fix the machine or close up shop.

How do you find out? You make a machine-capability study. At specified times during the day, take short consecutive runs of, say, five to ten parts and measure them carefully. Enter the results on an X-bar chart or into a computer, and find out whether the machine is in control. Once you establish this simple procedure, you can use it to check your machines in general, and to minitor on-going production jobs. That's how you kill two birds with the same procedure.

The people at Metal Seal know that discipline is more important than knowledge in making this sort of thing work. Top management must support it. Half of the seminar deals with management and how to set up an organization that can handle the program properly. When you make the magic step from QC to QA, you are well on the way to understanding the basics of SPC. Once you get the measurements taken on schedule, entered in the computer, or plotted on charts in good time, you can indeed control the process so you make few if any bad parts. The heart of SPC is feedback. You need it with your people (quality circles) as well as with your machinery and controls.

Photo: 1. Overview of Air Gage Co computerized gaging system for Ford Motor Co. Hydraulic pump is at left, not shown. Computer system is in protective cabinet at right.

Photo: 2. The workpiece is a transaxle case for front-wheel-drive cars. CAG system inspects every detail at end of line--fast enough for SPC. Manual in-line gages still serve at various points along the machining line to check special dimensions.

Photo: 3. Probe-spindle assembly with single probe. For the Ford setup, there are 88 probes in 18 spindles that make 83 inspection checks simultaneously. To prevent damage, the pallet fixture and gage spindles are equipped with air sensors.

Photo: 4. Dactron 1 signal conditioner and A/D converter provides probe excitation and signal demodulation in converting to a digital signal. The digital signal is then ready for processing, which will consist of either retention by the Dactron 1 for future interrogation, or transmittal of raw data to the computer for preparation of quality reports and statistical analysis.

Photo: 5. Gaging fixture at Station 5 in 11-station computerized in-line automatic system for checking conn rods.

Photo: 6. Cross Co installation of monitoring system provides statistical data for improved uptime.

Photo: 7. Mitutoyo MUX-40 interface system gets data moving fast from manual gaging directly to SPC computer.

Photo: 8. Brown & Sharpe PASS system serves SPC systems.

Photo: 9. DataMyte offers FAN (Factory Area Network) line of SPC equipment for easy data entry and computerized construction of control charts.
COPYRIGHT 1985 Nelson Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1985 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:statistical process control
Author:Miller, Paul C.
Publication:Tooling & Production
Date:Jun 1, 1985
Previous Article:Elements of machine-tool leasing.
Next Article:Quality wars: a report from general Juran.

Related Articles
Quality parts are as simple as SPC.
SPC software: ready when you are.
Combine training and SPC for quality improvement.
SPC for small to medium foundry's sand system.
Amoco Fabrics & Fibers: a quality journey of continuous improvement.
Under the spell of the quality gurus.
Certification is "no defects." (quality control standards)
'CQI' software goes beyond SPC.
Quality Assurance Revisited.
Pre-reflow, inline, 3-D inspection: the smart choice for your application?

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters