NMFRC
 

NCMSCompliance Assistance Centers

Funded by EPA
through a Cooperative Agreement

EPA

Disclaimer
The information contained in this site is provided for your review and convenience. It is not intended to provide legal advice with respect to any federal, state, or local regulation. You should consult with legal counsel and appropriate authorities before interpreting any regulations or undertaking any specific course of action.

Historical Articles

August, 1953 issue of Plating

 


Statistical Quality Control—A New Tool for the Electroplater

By Ezra A. Blount, Editor, Products Finishing Magazine


This paper was presented at the 40th Annual Convention of the American Electroplaters’ Society, June 15, 1953.

ABSTRACT
Statistical quality control, originated some thirty years ago as a method of inspection through scientific sampling to control quality in the mass production of machine parts to close tolerances, has been applied recently in the electroplating and finishing industries to control process variables such as solution composition, pH, and also to record and control the percentage of rejected parts. By using control charts for certain plating solution variables the operation can be controlled to a closer degree and a saving of plating chemicals can be made. Plated product inspection chars prove the value of close control of plating solutions. Adoption of these systems can reduce rejects and save plating metal. Examples of methods used in two large plating installations are cited.,

INTRODUCTION
At the 1951 American Electroplaters’ Society Convention held in Buffalo, N. Y., Dr. William Blum, in his keynote address, stated that appreciable further progress in electroplating research would require new tools for the researcher. The same situation applies generally to the electroplating industry, in that new equipment, new processes, new materials—in short, ”new tools”—are a continuing requirement. One such new tool, which the plating plant supervisor can employ as an aid in controlling and improving the quality of the plated products, is called ”statistical quality control”. It deals, essentially, with the inspection of manufactured products or processes, and involves the use of sampling techniques, as well as the laws of mathematical probability in determining mathematically certain probable or ”to be expected” variations. It may be described as the heart of scientific manufacturing, the basis for specifications, the means for manufacturing control, inspection, and quality certification.

Progress in industrial or scientific developments has generally followed two avenues of approach. Some discoveries have resulted from intensive research work with the specific purpose of proving an idea or theory, or creating new products or processes. Other equally remarkable developments have resulted from the application or adaptation of ideas or laws of one brand of science or industry to another entirely unrelated branch. The introduction of statistical quality control to the electroplating industry may be classified as an example of the latter type.

STATISTICAL QUALITY CONTROL—HISTORICAL BRIEF
Professor S. B. Littauer of Columbia University has traced the development of statistical quality control in the United States1, and described the establishment of an Inspection Engineering Department in the Western Electric Company to examine and interpret production inspection data. One of the first men assigned to this group was Walter A. Shewhart, who almost immediately introduced control charts as means of recording inspection data and controlling variations in various product characteristics. He then began his systematic development of the theory and practice of statistical quality control which was well rounded out by 1929. Two books by Shewhart record his work2 and assist in supporting the general consensus that Shewhart should be credited with the inception and development of the idea of statistical quality control.

An important part of the work on quality control carried on at Western Electric during the years 1924 to 1929 was the development of sampling inspection tables by Harold F. Dodge and Harry G. Romig, under the direction of George D. Edwards3. The first set of Dodge-Romig lot tolerance tables was completed in 1926, followed in 1927 by the first average outgoing quality limit tables.

In order to interest engineers as well as statisticians in industrial quality control, there was formed—with Mr. Shewhart as chairman—the Joint Committee for the Development of Statistical Applications in Engineering and Manufacturing, sponsored by the American Society of Mechanical Engineers and the American Society for Testing Materials and Joined by the American Statistical Association, the Institute of Mathematical Statistics, and the American Institute of Electrical Engineers. Later a Technical Committee on the Interpretation and Presentation of Data was formed under the sponsorship of A. S. T. M., and the work of this committee led to the publication of the first A. S. T. M. Manual on the Presentation of Data4.

General L. E. Simon, Ordnance Department, U. S. Army, began work on quality control in 1934. His work was valuable not only for the contributions he made to the practice of quality control5, but also because he was instrumental in getting the armed forces to promote nationwide use of statistical control in the procurement program. Civilian and military agencies, working together during the war years, spread the gospel of statistical quality control throughout American industry. An example of the continuing interest of military agencies in quality control may be cited in the presentation of a paper, ”Practical Implications of the Air Force Quality Assurance ;program”, by Lt. Col. O. C. Griffith, USAF, at the recent Seventh Annual Convention of the American Society for Quality Control.

During the past few years the idea of applying statistical quality control principles and methods to the electroplating industry has found increasing acceptance. For example, in their study of current distribution in barrel plating, Geissman and Carlson6 employed statistical quality control methods, and reported their findings at the 39th Annual Convention of the A. E. S. This study involved the measurement of deposit thickness, an important product characteristic, and was conducted to determine the average thickness of plate necessary to insure that a predetermined percentage of all parts would be plated to a thickness of deposit above the minimum specification.

STATISTICAL QUALITY CONTROL—APPLIED
Most electroplaters and chemists working in electroplating plants keep routine records of various chemical constituents or solution variables, such as pH, metallic nickel, boric acid, chloride, free cyanide, copper, chromic acid, sulfate ratio, depending upon the plating solutions used in a particular plant. Briefly, statistical quality control just carries these efforts a step further, and establishes the procedure on a firm scientific basis so that certain control factors may be calculated mathematically and variations in solution or operating conditions, greater than calculated allowable variations, may be noted and investigated.

Chemical additions made to a cleaning solution or plating bath under statistical quality control may be less or greater than additions made prior to its adoption. If solutions are permitted to deteriorate seriously before additions are made, then material costs under a system of statistical quality control will be less. However, it is more usual to expect solution additions to be slightly greater when operating under statistical quality control because the solution is kept up to proper operating strength at all times. However, the additions ill be uniform, (and will be made more frequently) and each addition generally smaller even though the monthly total may be more.

Fig. 1. Portions of two statistical quality control charts recording concentrations of alkali cleaner as well as daily additions to the cleaner tank. On both charts upper and lower limits of solution concentration, as outlined in the process specification, are indicated on the charts by horizontal dash lines. The upper chart shows daily concentrations before statistical quality control was installed in this plating plant. The lower chart shows the record of solution concentrations determined daily after statistical quality control methods were adopted. The conditions shown by the lower chart are much more-satisfactory because the cleaner concentration was kept within the specification limits at all times. At the same time a substantial material saving was accomplished (Illustration courtesy Ternstedt Division, General Motors Corporation, Detroit, Mich.)

The important advantage gained through statistical quality control of process variables is not a reduction in material cost, but a greatly lowered unit cost of production made possible by higher current densities, greater hourly production, and less defective work.
How charting the results of regular, daily chemical analyses of plating or other solutions and recording additions made can reduce material costs is shown in Fig. 1. Upper and lower control limits, as established by the process specification, are drawn on the chart, and then the solution concentration, in this case, oz/gal of alkali cleaner, is plotted on the chart each day— the sample of cleaner being taken and analyzed at’ the same time each day. The pounds of cleaner added each day are entered on another part of the same chart, so that the process supervisor can tell at a glance the solution concentration for any particular day, and the pounds of cleaner added on that day. A record of events or operating conditions which might bear on cleaner operation should be entered on the chart— much as daily events are entered in a ship’s log. Average cleaner addition prior to chart control was 183 pounds per day. After chart control was established, the average addition was 125 pounds per day, a 31 per cent saving.

The problem of inspecting the thousands of small parts or products that can be produced hourly from the larger automatic plating installations is no small one. While it may appear desirable to examine each part visually to be certain appearance standards are met, it is an almost impossible task to determine plating thickness, for example, on each part. Sampling inspection procedures may be employed effectively in this and other similar situations. If the per cent of defective or sub-standard items is below a certain value, then the lot is accepted. If the per cent defective exceeds the control limit, then every part in that particular lot is inspected. The saving in time made possible by sampling inspection is evident, both in the actual time required to perform the inspection, and in the elapsed time required for a certain number of parts to pass through this stage of the manufacturing process.

The effectiveness of statistical quality control procedures applied to plating processes in reducing rejects can be determined through inspection charts. The chart used to apply statistical quality control to in-process or final inspection of polished or plated parts or products is the p type of control chart, where-p is the fraction defective of parts from any one lot, and p is the average fraction defective from the entire inspection. Data concerning the lot number, sample size. (n), number of defective parts (pn), and fraction defective (p), are listed in table form. The average fraction defective (p) is calculated, and, again using standard mathematical formulas, the upper and lower control limits (UCL and LCL) may be calculated if desired. This information is plotted in chart form as illustrated in Figs. 2 through 4. Lots of parts, used as a basis for these data, have been inspected, presumably, by standard sampling methods to determine the fraction defective. Those lots with a fraction defective falling outside the control limits on the chart, can be set aside for 100 per cent inspection. When the fraction defective falls below the lower control limit an even more careful process investigation is made to determine what was responsible for the excellent results obtained. As the entire operation is brought under statistical quality control, however, the number of lots which require 100 per cent inspection will be greatly reduced, and, as shown on Figs. 2, 3, and 4, the average fraction defective is reduced. At the same time, the upper control limits are lowered. This results in proved, incontestable, increased quality at lowered production cost (material and time).

Fig. 2. Portions of charts upon which are plotted the percentage of copper plated parts rejected at the time of inspection. On the same chart the number of parts rejected for certain specific defects may be recorded. The upper chart shows the results of inspection for a three-week period prior to the introduction of statistical quality control to the plating process. As can be noted, the-average percentage rejection for that inspection period is 10.4. The lower chart records rejected parts over a similar period of time after statistical quality control was installed. Improvement in quality is indicated by the fact that the average percentage rejected during this period was 1.7. This result was accomplished in the main by improved line inspection procedures and by establishing and maintaining close controls-over the copper plating process (Illustration courtesy Ternstedt Division, General Motors Corporation, Detroit, Mich.)
Fig. 3. In the event that one particular defect (as shown by data recorded on charts similar to that shown in Fig. 2) indicates need for corrective action or more thorough investigation, that defect itself may be charted over a reasonable period of time until the defect is minimized or eliminated. In the portions of two charts shown, defects by reason of rough copper plate are recorded. The upper chart, showing conditions prior to the installation of statistical quality control methods, indicates that on the average 22.2 per cent of the parts rejected were rejected because of plate roughness. By closer control of the plating process and by certain adjustments made on the automatic plating equipment the average fraction rejected due to rough plate was reduced to 0.3 (lower chart) (Illustration courtesy Ternstedt Division, General Motors Corporation, Detroit, Mich.)

Fig. 2 is a comparison of copper plated and buffed parts rejected before quality control was installed, and afterward. The average fraction defective (p) was reduced from approximately 10 to about 1.6 per cent. The number of rejects from any one cause, such as rough plate, can be examined in the same fashion, and Fig. 3 shows how p for that defect was reduced from 22 to 0.3 per cent after statistical quality control methods were applied to the plating and polishing operation. A further proof of improvement at a point in the process where improvement is most noticeable can be seen in Fig. 4. These charts show that p, the average percentage of chromium plated parts rejected, was reduced from 14.7 to 4.6 per cent total rejects from all causes. Causes of rejects are listed on the same chart, to aid the process engineer in his continual fight to improve the process.

Fig. 4. A comparison of inspection charts showing results of inspection of chromium plated parts prior to and after the installation of statistical quality control methods in the plating operation. The average fraction rejected was reduced from 14.7 to 4.6. Also, as can be noted, the variation in the rejection percentage from the average has been greatly reduced. Improved fabrication and line inspection methods, as well as closer control of plating process, are credited with this improvement. (Illustration courtesy Ternstedt Division, General Motors Corporation, Detroit, Mich.) Fig. 5. Statistical quality control, to be completely effective, must keep supervisors and operators alike informed of the status of the various operations with which they are concerned. How this result is accomplished at Ternstedt Division, General Motors Corporation, is shown dramatically in this photograph. All the various charts applicable to the plating or inspection operations in a certain plant area are posted on bulletin boards readily accessible to all personnel concerned. Unsatisfactory conditions as well as improved conditions are immediately obvious. Thus, all can cooperate to eliminate unsatisfactory conditions and in the same way, all can feel a sense of pride and of accomplishment when the charts indicate improvement in quality (Illustration courtesy Ternstedt Division, General Motors Corporation, Detroit, Mich.)

SELLING STATISTICAL QUALITY CONTROL
The benefits resulting from the adoption of quality control procedures cannot be secured merely by training the plater, chemist, or process engineer in the use of the various charts, mathematical formulas, and procedures, and purchasing the required quantities of paper and ink with which to make the charts. It is necessary not only to train key individuals in the use of quality control procedures, but even more important, it is essential to interest all individuals concerned in the value of quality control, and create a desire on each person’s part to cooperate in making the project a success. Hence, a thorough educational program must be initiated. Each employee in the plating or polishing department must be made conscious of this new way of measuring results from a production operation, and must be awakened to its importance and value to him. Only after this spirit of cooperation has been established, can the system become truly effective.

Fig. 6. In one plating plant of the International Harvester Company cotton picker spindles are chromium plated on an automatic conveyor. One hundred samples are inspected from each hour’s production and the total number of rejects is recorded on the statistical quality control chart shown at the left in this photograph. As in the case of the charts illustrated in Figs. 2, 3, and 4, rejected parts are listed by cause at the bottom of the same chart so that an excessive number of rejects for any one particular cause may be investigated immediately and the unsatisfactory condition corrected. If the percentage of rejected parts is greater than 2 per cent for each hour’s production, then a 100 per cent inspection of that hour’s production is made. Normally the average number of rejects experienced is approximately 0.25 per cent, substantially below the upper control limit and far below the figure recorded prior to the installation of statistical quality control methods in this plating plant This control chart is posted immediately adjacent to the unloading end of the chromium plating machine and the inspection station so that inspectors, operators, and supervisors are made immediately aware of unsatisfactory conditions in the plating operation (Illustration courtesy International Harvester Company, Evansville Works, Evansville, Ind.)

INTANGIBLE BENEFITS FROM STATISTICAL QUALITY CONTROL
As can be noted in Figs. 5 and 6, quality control charts are posted in a conspicuous location immediately adjacent to the plating operation, or point of inspection, so that the operators and supervisors alike can see improvements in quality as reflected in the various charts. They are thus continually reminded of the value of statistical quality control. Such a program has the further value of dramatizing the position and importance of many employees, who might previously have had the feeling that they were merely parts of a production machine. As everyone knows, an interested employee, and an employee who is made to see and feel his importance, is a much more valuable employee.

Perhaps the improvement in production resulting from improved employee morale—particularly in the larger plating shops—is even greater than that resulting from material and time saving made possible through statistical quality control. In any case, all these benefits are being obtained in the plants where statistical quality control has been properly installed, and is being used properly.

ACKNOWLEDGMENT
The author wishes to express his deep appreciation to Bryant W. Pocock and Frank L. Bonem, consulting and contributing editors of ”Products Finishing”, for their assistance in assembling the material on which this paper is based.

REFERENCES
1. S. B. Littauer, ”The Development of Statistical Quality Control in the United States”, published in the December, 1950 and the February, 1951 issues of The American Statistician.
2. W. A. Shewhart, ”Economic Control of Quality of Manufactured Product”, D. Van Nostrand & Co., Inc., 250 Fourth Avenue, New York. N. Y., 1931. ”Statistical Method from the Viewpoint of Quality Control”, (Edited by W. Edwards Deming) The Graduate School, Department of Agriculture, Washington D. C., 1939. ‘
3. H. F. Dodge and H. G. Romig, ”Sampling Inspection Tables —Single and Double Sampling Tables”, John Wiley & Sons, Inc., 440 Fourth Avenue, New York, N. Y., 1944.
4. A. S. T. M. Manual on Presentation of Data”, American Society for Testing Materials, 1916 Race Street, Philadelphia 3, Pa.
5. L. E. Simon, ”An Engineer’s Manual of Statistical Methods” John Wiley & Sons, Inc., 440 Fourth Avenue, New York, N. Y., 1941.
6. W. G Geissman and R. A. Carlson, ”Current Distribution in Barrel Plating”, Proc. Am. Electroplaters’ Soc. 39, 163-165 ( 1952).
7. E. L. Grant, ”Statistical Quality Control”, McGraw-Hill Publishing Co., Inc., 330 W. 42nd Street, New York, N. Y., 1946.
8. A. V. Feigenbaum, ”Quality ControUPrinciples, Practice and Administration”, McGraw-Hill Publishing Co., Inc., 330 W. 42nd Street, New York, N. Y., 1951.
9. E. M. Schrock,”Quality Control and Statistical Methods”, Reinhold Publishing Co., 330 W. 42nd Street, New York, N. Y., 1950.
10. N. L. Enrick, ”Quality Control—Manual of Practical Pro” cedure for Shop and Plant Operation”, The Industrial Press, 148 Lafayette Street, New York, N. Y., 1948.

Discussion
MR. H. A. FUDEMAN (Trico Products Corporation, Buffalo, N. Y.): Unfortunately, plating inspection still relies on the human eye. Do you know of any devices in use that would replace this method?

MR. BLOUNT: There are some devices in use which measure appearance characteristics, depending upon the properties that you wish to inspect for, but generally in inspecting for appearance, the human eye is still the fallible though useful tool.

DR. A. M. MAX (RCA Victor, Indianapolis, Ind.): I would like to make a comment that physical inspection does not reduce rejects. Statistical quality control is simply a tool to assist in product control and remove the causes for rejects.

MR. BLOUNT: Thank you very much, Dr. Max. Yes, I am familiar with the fact that you cannot inspect quality into the product, you have to build it in.

MR. WILLIAM GEISSMAN (National Lock Company, Rockford, Ill.): I certainly want to again impress upon you the statement the author made when he said that you have to begin at the beginning. You must obtain the whole-hearted cooperation of your entire plant in order to use this tool called Quality Control. In other words, one cannot expect you to put on a specified plate thickness of, for instance, 2 ten-thousandths of an inch and then have the part which came to you undersize by much more than that and then build it up to the proper size. Neither can you plate parts that are oversize before plating and expect them to be within tolerance after plating. These parts should have been rejected before plating. Nor can you level out gross polishing marks or tool marks and expect to get a good looking part. Here again rejection before plating is a must.

I also have a question for the author. Are any companies using statistical methods to evaluate the appearance of plated ware for decorative purposes? We know these methods are being used in the paper industry and the leather industry, but as yet we have not found anybody using it in the plating industry. We are trying to work out a method ourselves, but it is a long, hard row.

MR. BLOUNT: I don’t know anything about the evaluation of appearance in the plating industry, either, Mr.. Geissrnan. You are probably familiar with the work being done in the porcelain enamel industry along that line; they have evaluated various surface defects. If you study that perhaps you can borrow some ideas from another industry.

MR. FRANK TIRENDI (Patent Button Company, Waterbury, Conn.): How is the upper and lower control limit determined for a given process?

MR. BLOUNT: In two ways—one by the process specification which the plating solution supplier or your own process engineer establishes; the other, by means of the statistical control method. Upper and lower control limits are calculated by two relatively simple formulas which I did not go into because I wanted to keep mathematics out of the discussion. They are discussed in the A.S.T.M. Manual on ”Quality Contro] of Materials”.

MR. THOMAS J. MENZEL (Jack Steele Company, Philadelphia, Pa.): It is true, Mr. Blount, that analysis is of prime importance. However, I do want to suggest that analysis complemented by plating tests is true control.

MR. BLOUNT: Thank you, Mr. Menzel.

MR. WILLIAM E. TREMBLEY (Texas Instruments, Inc., Dallas, Texas): By statistical control do you mean the installation or use of in-process inspection to a greater degree than say we have it at present?

MR. BLOUNT: That is correct. It is not only in process inspection, but control of the product all the way through the process. The application of these charts and methods carries process inspection one step farther.



| Compliance Assistance | Regulations | Directories | Resources | Hot Topics | News | Ask the Experts | Library | Online Training | About NMFRC | Search | Home |

NMFRC