ASM Metals HandBook Vol. 14 - Forming and Forging

Подождите немного. Документ загружается.

Fig. 13

Measurements from two points in time (values A and B) would suggest that material B has the lower

forming severity. However, when full control charts are evaluated, material A has the lower severity.

If the forming system is unstable, then the system must first be stabilized before any changes are made and documented.

The following example illustrates how statistical deformation control can be used to identify the sources of processing

difficulties.

Example 6: SDC Analysis of an Outer Side Panel of a Truck Body.

The production of a truck outer side panel was marked with sporadic breakage problems. First, high breakage rates were

recorded in area A of Fig. 14 for limited but specific lots of steel brought to the press. Mechanical-property evaluation of

these high-breakage lots showed certain formability parameters to be equivalent to those measured in other lots of the

same steel type that did not break. Second, certain extended runs of the part experienced more incidents of breakage than

other runs.

Fig. 14 Illustration of truck body side panel showing critical-strain area A and zero-

strain area B used for

ultrasonic thickness measurements.

Process Analysis. This problem was selected as a trial for statistical deformation control. The following six steps were

utilized.

First, a circle grid analysis of the stamping was performed using a sheet of the current production material that was

satisfactorily making the stamping (Ref 8, 9, 12). The most severe strain was located in area A of Fig. 14. Plotting this

strain state on the forming limit diagram showed a safety margin of 8 strain%. This means that the stamping had marginal

severity but would not be expected to fracture if the current conditions of the forming system were maintained.

Second, the circle grid analysis showed no strain in area B of Fig. 14. This location was designated as the as-received

sheet thickness for purposes of the ultrasonic thickness measurements. This location was ideally suited to thickness

measurement because it was at the same relative edge-center-edge location in the coil as the most severe strain location.

Third, a template was prepared for the stamping by cutting a section of another stamping to encompass the two locations.

A hole was drilled at each location to accommodate the head of the ultrasonic probe.

Fourth, one lot of the production steel that was successfully making the stamping was set alongside the press as a

reference lot. Blank edges were marked with yellow paint to identify at the end of the line all stampings made from this

reference lot.

Fifth, at the halfway point of each lot of production steel brought to the press, five of the reference blanks were formed.

Stampings from these blanks, plus the next five production stampings, were removed from the end of the production line

for evaluation. Thickness ratio t

was determined for each stamping. The

and R values were calculated for both the

production steel and the reference steel.

Sixth, concurrent plots of the

and R values for the production and the reference steel lots were maintained. Sections of

the

plot are shown in Fig. 15. The mean, the upper control limit, and the lower control limit are not shown, because the

solution to the problem was achieved before sufficient data points were collected for these statistical calculations.

Fig. 15 Data for two days of measurements made on the truck body side panel shown in Fig. 14.

See text for

details.

Solution. The left-hand side of Fig. 15 verifies the initial circle grid evaluation that the stamping has some safety margin

but that the safety margin is insufficient; the value of the thickness ratio is located in the bottom half of the marginal zone.

In this case, it is possible to evaluate the severity of each reading as it is made; the quantity of data points needed to create

a control chart is not required in order to determine the forming severity of each stamping.

Both the production steel and the reference steel appear to be maintaining not only a constant but also an equal forming

severity. Neither the production steel nor the forming process provides indications of impending breakage.

One lot of steel that was previously rejected for high breakage rates (retry lot B, Fig. 15) was retried; a marked increase in

forming severity to the failure zone was noted. The formability of sheet steel is an interaction of substrate properties and

the lubricity of the steel/lubricant combination. However, this lot of steel had forming properties equal to or better than

the production and reference steels noted in Fig. 15. Therefore, the lubricity of the steel surface was suspected; the steel

was coated on the punch side of the sheet with a corrosion-resistant paint. To test the influence of the painted surface,

several blanks were turned over so that the bare side of the sheet contacted the punch. These sheets were successfully

formed into stampings without breakage. Although unacceptable for production in this reversed condition, these

stampings verified that the substrate had sufficient formability to produce the stamping without breakage when the

interface friction was correct.

It was recommended that the painted (punch) side of the blanks be sprayed with lubricant if breakage was encountered

with a specific lot of steel. If breakage persisted with the additional lubricant, the lot was to be rejected for laboratory

analysis of the substrate properties.

The right-hand side of Fig. 15 illustrates another problem. When the tooling was reset in the press for another run,

extensive breakage was encountered with many of the lots of steel brought to the press. An SDC analysis of the stamping

showed that the severity of the stamping for the reference steel had increased to zero safety margin, which meant that the

tooling was not performing identically to the previous run. This indicated a different setting in the press or modifications

made to the tooling during the period the tool was out of the press. Discussion with the tool and die staff revealed that the

draw beads and binder surfaces of the tooling were reworked in an attempt to reduce low spots in the stamping. This

reduced the flow of metal from the binder areas and increased the stretch component required over the punch, thus

increasing the stamping severity. The increased severity was verified when samples of the previous problem lot B were

inserted in the tooling. Long splits were encountered compared to the previously encountered slight necking condition.

The recommended solution was to encourage additional metal flow from the binder area through reduced binder pressure

until the reworked draw beads became reduced in effectiveness through normal wear.

Production steel E had a slightly increased severity as compared to production steels A, C, and D. However, production

steel E had a lower severity than the reference steel; production steels A, B, and D had a severity comparable to that of the

reference steel. Using the reference steel as a common base for comparing various production runs, steel E actually had

better formability than steels A, B, D, and the reference steel for the specific stamping under investigation.

Continued collection and monitoring of data during the entire production run was recommended. This would require a

number of reference lots. However, the SDC procedure allows for this transition by the interleaf forming (alternate

blanks) of both reference steels to calibrate the new reference lot relative to the previous reference lot.

References cited in this section

1. Statistical Quality Control Handbook, Western Electric Company, 1977

2. J.M. Juran, F.M. Gryna, Jr., and R.S. Bingham, Jr., Quality Control Handbook, McGraw-Hill, 1979

3. E.L. Grant and R. S. Leavenworth, Statistical Quality Control, 5th ed., Mc-Graw-Hill, 1980

4. A.L. Strongrich, G.E. Herbert, and T.J. Jacoby, "Statistical Process Control: A Quality Im

provement Tool," Paper

83099, Society of Automotive Engineers, 1983

5. General Motors Statistical Process Control Manual, Document 1893, General Motors Corporation, 1984

6. J.F. Siekirk, Process Variable Effects on Sheet Metal Quality, J. Appl. Metalwork., Vol 4 (No. 3), July 1986, p 262-

269

"Final Report on Manufacturing Cost Study Conducted by Pioneering Engineering," American Iron and Steel

Institute, 1986

8. S. Dinda, K. James, S. Keeler, and P. Stine, How to Use Circle Grid Analysis for Die Tryout,

American Society for

Metals, 1981

9. Sheet Steel Formability, American Iron and Steel Institute, 1984

10.

S.P. Keeler, Understanding Sheet Metal Formability, Sheet Met. Ind., Vol 48 (No. 5-10), 1971

11.

W.G. Brazier and S.P. Keeler, Relationship Bet

ween Laboratory Material Characterization and Press Shop

Formability, in Proceedings of Microalloying 75, Union Carbide, 1977, p 517-530

12.

S.P. Keeler, "Statistical Deformation Control for SPQC Monitoring of Sheet Metal Forming," Paper 850278, Society

of Automotive Engineers, 1985

Statistical Analysis of Forming Processes

Stuart Keeler, The Budd Company Technical Center

Experimental Design

Forming operations span an extensive production life cycle, including design, prototyping, engineering changes, tryout,

and production. During this entire cycle, a pervasive question is whether the process can be modified. Process

optimization is the usual motivation for this modification. This optimization may mean lower cost, faster production rates,

fewer rejects, better quality, and so on.

Some process simulation techniques may be available for determining the steps needed to optimize the process and the

effectiveness of the various options. However, most of these simulation techniques are mathematically based and only

approximate the system under evaluation (Ref 13, 14, 15). Therefore, most optimization studies are conducted on the

system itself by selectively changing the system variables and measuring the effects of the changes.

These system studies are most effectively conducted with experimental design. Experimental designs are used to increase

the efficiency of information acquisition.

Several forms of experimental design are available, and each has advantages and disadvantages. The types of experiments

discussed in this article are single-variable studies, multi-variable studies, and Taguchi experiments. More information on

the design and execution of such experiments is available in the article "Process Modeling and Simulation for Sheet

Forming" in this Volume.

Single-Variable Studies

The most common experiment conducted with forming operations is the single-variable study. Although many variables

are identified as affecting the process, only one of these variables is changed at a time.

Single-variable studies are widely used because they are easy to conduct and require a minimum number of experiments.

First, one variable is changed, and then another, and so on, until all of the variables are at the second level. One such

experimental design table is shown in Fig. 16. Here the (+) indicates the high level of the variable, and (-) indicates the

low level of the variable. The following example outlines the procedure used in a typical single-variable experiment.

Fig. 16 Layout of an experimental design for a single-variable investigation.

Example 7: Finding Improved Process Settings for the Production of an Automobile

Inner Door.

High breakage rates were observed during the die tryout of an automotive inner door. A quality group met and identified

seven key factors (controlling variables) that affect this particular forming process.

Two steels and a lubricant were obtained. First, the stamping was made with steel A; all other variables were kept at their

existing conditions. The severity of the process was determined using the thickness ratio analysis described above. The

stamping was then made using steel B.

Next followed the additional six changes indicated in Fig. 16. The thickness ratio was determined for each change. After

all eight experiments were conducted, the level of each variable showing the greatest thickness ratio (least amount of

metal thinning) was selected as the new standard operating procedure.

Problems With Single-Variable Studies. In this study, one variable was changed while all other variables were

kept constant. The problem is that the conclusions obtained from this experiment are valid only if all other variables are

kept identical to the experimental value. This type of experiment ignores synergistic effects among process variables. In

the experiment described in Example 7, the effects of different materials (steels A and B) can be measured only if variable

C (lubricant) is maintained at its high level (with lubricant). If variable C is at its low level (no lubricant), misleading

results may be obtained. In this case, one cannot deduce what variable A does for all levels of the other variables.

Therefore, some type of factorial experiment (orthogonal array) is required.

Multi-Variable Studies

Multi-variable studies are usually conducted using some form of an orthogonal array. The experimental layouts using

these orthogonal arrays yield experimental results such that the effects of varying a given parameter can be separated

from other effects. Figure 17 shows a complete orthogonal array, sometimes called a full factorial experiment, for seven

variables. These orthogonal arrays have also been termed square games (Ref 16).

Fig. 17 Layout of a full factorial (multi-variable) experimental design for a group of seven variables.

The term orthogonal means balanced, separable, or not mixed, and this indicates that each variable is evaluated equally

for all other conditions. Therefore, for each of the two levels of any variable in the array in Fig. 17, there are 64

combinations of the other variables. For example, for level A

in Fig. 17, there are 64 possible combinations of variables

B, C, D, E, F, and G. This also holds for level A

. Therefore, any effect of level A

versus level A

is determined in the

presence of both the high and low levels of all the other variables.

When the quantity of parameters is increased in these factorial experiments, the number of experiments required increases

so rapidly that it may not be feasible to implement the experimental design. The full factorial experiment detailed above

involves 128 experiments. Increasing the number of variables by 1 doubles the number of experiments required to 256.

The full factorial experiment provides the effect of all the single-variable effects plus all the interactions. If some of the

interactions can be ignored or deemed unimportant, then the experiment can be redesigned as a partial factorial. A partial

factorial has a reduced number of required experiments. The procedure used in one full factorial analysis is outlined in the

following example.

Example 8: A Full Factorial Experiment for Finding Improved Process Settings for

Stamping an Automobile Inner Door.

The same quality group that conducted the experiment described in Example 7 realized that the synergistic effects were

not being evaluated with a single-variable experiment. The group decided to design a full factorial experiment in which

all interactions would be active in the study and all interactions could be calculated from the data obtained.

The full orthogonal array shown in Fig. 17 was used. The same variable assignments used in Example 7 (see Fig. 16)

were repeated in this study. Again, ultrasonic measurements of sheet thickness were used to determine the thinnest area

(maximum metal thinning) for calculation of stamping severity. Analysis of the results showed that some of the

interactions were very important and had to be considered in selecting the best operating levels of each parameter.

Taguchi Experiments

Conventional experimental design techniques were primarily developed for use in scientific research in order to determine

cause-effect relationships (Ref 16). In science, there is generally only one law to explain a natural phenomenon; therefore,

the primary experimental efforts are aimed at finding the law that explains the relationships being studied.

In technological fields, however, there are numerous ways to obtain a given product design objective. In forming a sheet

metal stamping, there are various steel characteristics, die designs, die steels and surface treatments, lubricants, forming

sequences, and other variables to be considered. Different combinations can be used to produce the same end design of

stamping. A desirable goal is to find the combination that provides the most stable and reliable performance at the lowest

manufacturing cost. This is the goal of the Taguchi experiments--to develop the most robust process possible (that is, the

process that is least sensitive to the causes of variation).

In many cases, knowing the cause of a problem is insufficient for solving the problem. Removal of the cause of the

problem may be too costly or even impossible. The Taguchi strategy then becomes one of finding countermeasures--not

to eliminate the cause but to reduce the influence of the cause on the end product. Efforts are aimed at making the final

product insensitive to all process variables. Taguchi techniques are oriented more toward cost effectiveness and marketing

than are conventional experimental design techniques (Ref 16). This difference affects the nature of the parameters to be

cited, the way in which the experiments are laid out, and the way in which the data are analyzed. In a sense, the Taguchi

method of conducting experiments is formalized to such an extent that only minimal exchange of information is required

among experimenters trained in this technique in order to achieve complete understanding of the experimental

parameters, the experiments, the analysis of the data, and the results/conclusions.

Differences between the Taguchi approach to experimental design and the conventional approach are philosophical and

methodological (Ref 16). Philosophically, the Taguchi approach is technological rather than theoretical, inductive rather

than deductive, an engineering tool rather than scientific analysis. Taguchi emphasizes productivity enhancement and cost

effectiveness with a rapid, formalized procedure--not statistical rigor. With regard to its methodology, the Taguchi

approach emphasizes the application of orthogonal arrays, signal-to-noise ratios (in Taguchi analysis, the variables that

affect a process are classified as signal, control, or noise factors), and newly developed analytical techniques (Ref 16).

One primary advantage of the special orthogonal arrays used by Taguchi is the capacity and flexibility of assigning

numerous variables with a small number of experiments. An even more important advantage of the array is the

reproducibility of the conclusions across many different process conditions. One criticism is that the Taguchi array is

nothing more than a fractional factorial, which has long existed in conventional experimental design techniques.

However, the Taguchi technique is one of philosophy and content, as explained above. The application is more versatile

and sophisticated.

Further explanation of the Taguchi method or the experimentation/analysis procedures is beyond the scope of this article.

Additional information is available in Ref 16 or from the American Supplier Institute (Dearborn, MI). The following

example outlines the approach used in one Taguchi analysis.

Example 9: A Taguchi Design of Experiments for Finding Improved Process Settings

for Stamping an Automobile Inner Door.

The same seven variables used in Examples 7 and 8 are used in this experiment. Selection of the proper experimental

variables is extremely important in Taguchi experiments. The variables should be primary variables with little interaction.

In studying stamping severity, for example, blankholder pressure (variable F) is not set by the tons of force shown on a

load indicator, because the measured tons are highly interactive with all of the other variables of the system, such as blank

thickness, metal strength, die temperature, and shims used. Instead, the blankholder pressure variable is controlled by the

position of the nut (or turns of the screw) on the outer ram relative to the zero load reference position. The extra time

spent identifying and selecting the important process variables is the key to the success of the Taguchi experiment.

Figure 18(a) shows the layout of the Taguchi experiment for the seven variables detailed in Example 7 (Fig. 16). The

array is a standard array for up to seven controlling variables or factors. The eight experiments required for a full factorial

analysis are shown in Fig. 18(b).

Fig. 18 Layout of the Taguchi experimental design for seven variables.

There are eight rows representing eight experiments, numbered 1 through 8. The elements of the seven columns consist of

1s and 2s. There are four 1s and four 2s in every column. In any pair of columns, there are four possible 1, 2

combinations; namely 11, 12, 21, and 22. Because each of these four combinations occurs at an equal number of times in

a given pair of columns, the two columns are balanced or orthogonal. Figure 18 shows that all column combinations have

an equal number of 11, 12, 21, and 22 combinations; therefore, all interactions are within the experimental design. After

the eight experiments are completed, the data are analyzed according to a prescribed procedure in order to determine the

level of each variable that contributes to the most robust (most insensitive process) combination of processing variables.

References cited in this section

13.

N.M. Wang and S. Tang, Ed., Computer Modelin

g of Sheet Metal Forming Processes: Theory, Verification, and

Application, The Metallurgical Society, 1986

14.

H. Yamasaki, T. Nishiyama, and K. Tamura, Computer Aided Evaluation Method for Sheet Metal Forming in Car

Body, in Proceedings of the 14th Biennial Congress of the IDDRG

(Koln, Germany), International Deep Drawing

Research Group, April 1986, p 373-382

15.

K. Chung and D. Lee, Computer-

Aided Analysis of Sheet Material Forming Processes, in Advanced Technology of

Plasticity, Proceedings of the First International Conference of Plasticity

(Tokyo), Japan Society for Technology of

Plasticity and Japan Society of Precision Engineering, 1984, p 660-665

16.

Y. Wu and W.H. Moore, Quality Engineering--Product and Process Design Optimization, American Suppl

ier

Institute, 1985

Statistical Analysis of Forming Processes

Stuart Keeler, The Budd Company Technical Center

References

1. Statistical Quality Control Handbook, Western Electric Company, 1977

2. J.M. Juran, F.M. Gryna, Jr., and R.S. Bingham, Jr., Quality Control Handbook, McGraw-Hill, 1979

3. E.L. Grant and R. S. Leavenworth, Statistical Quality Control, 5th ed., Mc-Graw-Hill, 1980

A.L. Strongrich, G.E. Herbert, and T.J. Jacoby, "Statistical Process Control: A Quality Improvement Tool," Paper

83099, Society of Automotive Engineers, 1983

5. General Motors Statistical Process Control Manual, Document 1893, General Motors Corporation, 1984

6. J.F. Siekirk, Process Variable Effects on Sheet Metal Quality, J. Appl. Metalwork., Vol 4 (No. 3), July 1986, p 262-

269

"Final Report on Manufacturing Cost Study Conducted by Pioneering Engineering," American Iron and Steel

Institute, 1986

8. S. Dinda, K. James, S. Keeler, and P. Stine, How to Use Circle Grid Analysis for Die Tryout,

American Society for

Metals, 1981

9. Sheet Steel Formability, American Iron and Steel Institute, 1984

10.

S.P. Keeler, Understanding Sheet Metal Formability, Sheet Met. Ind., Vol 48 (No. 5-10), 1971

11.

W.G. Brazier and S.P. Keeler, Relationship Between Laboratory Material Ch

aracterization and Press Shop

Formability, in Proceedings of Microalloying 75, Union Carbide, 1977, p 517-530

12.

S.P. Keeler, "Statistical Deformation Control for SPQC Monitoring of Sheet Metal Forming," Paper 850278, Society

of Automotive Engineers, 1985

13.

N.M. Wang and S. Tang, Ed.,

Computer Modeling of Sheet Metal Forming Processes: Theory, Verification, and

Application, The Metallurgical Society, 1986

14.

H. Yamasaki, T. Nishiyama, and K. Tamura, Computer Aided Evaluation Method for Sheet Metal Fo

rming in Car

Body, in Proceedings of the 14th Biennial Congress of the IDDRG

(Koln, Germany), International Deep Drawing

Research Group, April 1986, p 373-382

15.

K. Chung and D. Lee, Computer-Aided Analysis of Sheet Material Forming Processes, in Advance

d Technology of

Plasticity, Proceedings of the First International Conference of Plasticity

(Tokyo), Japan Society for Technology of

Plasticity and Japan Society of Precision Engineering, 1984, p 660-665

16.

Y. Wu and W.H. Moore, Quality Engineering--Product and Process Design Optimization,

American Supplier

Institute, 1985