Printer Friendly

Let's do it right.

Let's do it right

Have you ever tried to explain rubber testing to a person new to the industry? It's especially fun when what person is an irate customer who discovers that the tensile strength in a failed part is 50 psi less than a good sample.

It doesn't take too long in our business to discover that physical property data is almost always subject to interpretation. My personal rule for comparing data is the 10% rule - any numbers within 10% of each other are within the experimental error of the test itself.

My experience has been that this is a good rule. However, it also means that, in practical terms, it's difficult to say that a compound with a tensile strength of 3,000 psi is much different than one with a strength of 3,250 psi. Or 2,750 psi.

That's a wide range. Most rubber operations, including the major rubber houses, don't take the pains required to reduce it. However, it can be done.

The rest of this column will talk some about how more accurate test data can be obtained. Since tensile data is the most "artsy," I'll focus on it.


Companies will spend many thousands of dollars on new tensile testing equipment that is capable of measuring tensile strengths to an accuracy of 1 psi or elongations to 1%. Yet these same companies will often ignore a much greater source of error - how the sample is prepared.

Preparation starts with milling. The rubber sample is milled out to the desired thickness and cut to the appropriate size to be put in the mold. Unfortunately, the thought process often stops here.

Miling itself requires some definition. Rubber will be warmed up on the mill and a grain will be established in the material. The grain, of course, will run in the same direction as the mill runs.

In preparation of samples for tensile testing, the rubber should he warmed thoroughly. If the sample is blended before the final sheeting off, a standard method of blending should be established. For example, if the rubber is rolled off the mill and fed end-wise back into the mill, a more random grain will be achieved than if it is allowed to feed lengthwise back into the mill.

Mill thickness settings and mill ratio (relative speed of front to back roll) will also affect the establishment of grain in the sample. Mill thickness should be set so that the sheet taken off the mill is 0.002 to 0.0005 inch thicker than the cavity in the tensile mold. Since different types of rubber tend to swell differently and have different degrees of nerve, no exact setting is possible for all formulas. Typically, the mill opening will be somewhere between 0.060 and 0.080 inch.

Thickness control on the sample is important to limit flow in the mold. Flow of rubber in the mold will affect grain alignment which will affect test results. By reducing mold flow to a minimum, grain alignment established by the mill will be least affected. Grain direction should be marked on all samples as they are cut from the milled sheet. If the milled sample does not fill out the width of the mill (making grain direction obvious), direction should be marked on the sheet immediately after it is taken off.

Once the sample is milled, best (most reproducible) results will be obtained by allowing the rubber to thoroughly cool and rest. Recommended minimum time is four hours. This allows all the strained elastomer chains to relax into their most normal state. After the rest period is complete, curing can be done.


Curing again seems simple - and it is. But attention to detail (or lack of it) will have significant effects on physical property results.

First, the most obvious - cure temperature and time. Temperatures in lab presses should be controlled to +/-2 [Degrees] F or less. In some operations, I've seen presses that typically can't hold more than +/- 10 [Degrees] F. Even a fluctuation of 5 [Degrees] can result in one sample seeing effectively 50% more cure than another.

Electrically heated platens tend to have more fluctuation in temperature than a well controlled steam heated platen. Also, controls must be available so that in electrically heated platens with multiple elements, it is possible to detect when one of the elements burns out.

Steam, however, is not without its problems. Steam traps must be in excellent working order and steam channels must be clean and free from scale. If any water is allowed to condense in to the platen, it will create a cold spot on the platen.

When the rubber sample is placed in the mold, it is important that the grain direction from milling be marked in some fashion on the sample.

Cure times used must be consistent. For example, if a sample is to be cured seven minutes at 330 [Degrees] F, should the timer be set as soon as the sample is placed in the hot mold, as the press is closing or after any bump cycle is complete? All of these are acceptable procedures as long as they are done consistently. Otherwise, in the example just given, the cure time could actually vary as much as one minute - almost a 15% difference.

When the cure is complete and the timer goes off, the sample must he removed immediately from the mold and cooled. Immediate removal means that within 10 seconds of the timer going off, the mold is being opened and the sample removed. Cooling is best done in a cold water bath. Cooling immediately stops the curing process. If the sample is allowed to cool slowly, the curing process can continue for some additional time, again producing variable results.

Once the sample has been removed from the mold, it should be allowed to rest for at least 16 hours prior to testing. This again allows full relaxation of the compound. It also will produce the most consistent results. I have personal experience with compounds that have had property increases as much as 10% simply by letting them rest.


Nobody ever talks about cutting samples. Yet it is a very important part of the testing process. The most critical aspect of it is the condition of the die used. Tensile and tear dies must be level and sharp. Any nick present on the cutting surface can cause significant problems with results. Even brand new or resurfaced dies need additional honing after receipt. Normally this is done with a hard Arkansas stone type file.

Levelness of the die can be checked easily with a straightedge across the cutting surface. The edge can be felt to see if it is free of nicks and problems. It is also wise to examine it under a magnifying glass.

Finally, die condition can be checked by cutting out a sample and examining the cut edge of the sample. Any irregularities on the cutting edge will be easily observable on the edge when properly lighted.

Automatic punch type cutters work the best, removing operator variables from the cutting process. If the cutting is done manually with a mallet, it is important that the die is struck evenly and hard enough that a sample is completely and cleanly cut with one blow. In either case, the cured sheet from which the sample is to be cut must be evenly supported under the die. One of the best methods of doing this is with layers of smooth cardboard.

Unless testing for special properties, both tensile and tear sample should be cut so that the grain direction from the milling operation runs lenghtwise along the sample. This will produce the highest and most consistent values.

Testing machines

After all this, we can finally begin talking about the testing machines. There are many different types to tensile testers used in the industry today, ranging from the old pendulum type requiring elongation to be measured using eyeballs and a ruler to new, high precision load cell units that track elongation without touching the sample.

First, all of the machines will work to one degree or another. With a well trained, experienced operator, surprisingly accurate results can be obtained from the old pendulum type machines. One of their drawbacks is that no two operators can get quite the same results from the same sample. The problem largely comes from the fact that, having to use a ruler to tract elongation, no two people will follow the sample with the ruler quite the same.

There are also limitations on the accuracy of these machines' measurement of stress on the sample as it is pulled. Non-linear relationships between the application of load and measurement cause some of this.

All newer machines use load cells to measure the application of stress. This eliminates many of the variables. Also, most new machines have extensometers of some type available to measure strain as the sample is pulled. All of this improves the accuracy of the results. Differences between units come from accuracy of the load cell, accuracy of the extensometer, accuracy and repeatability of the rate of travel of the head of the machine, etc. Machines will also differ in their ability to achieve very high or low strain application rates and ability to perform more complex dynamic mechanical functions.

Regardless of the type of machine used, several factors will affect results. For example, samples should be "relaxed" when mounted in the machine. Preloading to any degree will cause erroneous results.

Likewise, if the sample has been deformed in any way prior to the test, its results will be invalid.

Most tensile molds produce a cured sheet approx. 0.070 inch thick. As the thickness of the sample increases, problems will arise from a dishing out of the sides of the sample during cutting. This can be easily seen in samples of soft compounds that are over 0.100 inch thick. This type of problem will cause erroneous test results. Unfortunately, there is no easy remedy except to reduce the sample thickness.

Another source of error is in measuring and recording the thickness of the specimen. Thickness measurements should be made in at least three locations in the web of the sample with the median being used for calculations. If the difference is over 0.003 inch, the sample should be discarded.

Final examination

Once the samples have been tested, the broken pieces should be saved and examined, particularly in the case of tensile specimens. The examination should determine:

* Did the samples all break in the same place?

* Were all the breaks on the same side of the test machine?

* Are there any odd characteristics about the breaks?

Once the data is available, how do you know it's "right?" Aside from regular calibration of the test machines, the best way is to periodically cross-check with someone else in the business. A simple check would be to send the died out samples for testing and compare their results to your own.

Another way is to subscribe to one of the test services available, such as SPControls. These people provide a subscription service which periodically sends out samples for testing. Your operator performs the tests, returns the results, and they collate your data with all others using the service and advise you of how your results compared to the rest of the population.

When close agreement is needed with a vendor or customer, it may be necessary to set up a designed experiment to fully define the relationship using standards produced by both parties.

There are advantages and disadvantages to all these techniques. The first method is the simplest and least expensive to perform. However, it limits the accuracy of the comparison (you are assuming the other guy is right).

The second method improves accuracy assuming that the median of the group is probably accurate. However, it will take a bit more time and will cost more. Both these methods have the additional limitation that they focus on the test machine and ignore all the preparation, curing and cutting steps.

The final method can be the most expensive. It doesn't worry about absolute right and wrong values as much as it establishes a relationship between two laboratories. Properly set up, it will define overall differences as well as specific differences between the two. Its drawback is that, since it cannot define absolutes, it is only useful to the two labs.


As we mentioned before, rubber is a bit of a black art. Not the least of the problem is simply the problems and potential for error involved in the testing.

While calibration using dead weight standards is necessary, it doesn't always give the whole story. I had one occasion in the past where a machine was used that had been faithfully maintained and calibrated for years. Everyone expected the results received from it to be accurate. However, when an outside crosscheck was performed, it was found that it produced tensile strength results 3-400 psi lower than other machines in the industry. No satisfactory explanation was ever found. Because of that difference, it was determined that, over the years, in designing compounds to meet certain specifications, they had been inadvertently overcompounded. As a result, compound costs were higher than they needed to be.

Having accurate test data requires looking at all aspects of the testing procedure. Having an expensive, fancy machine can be nice, but it cannot overcome poor sample preparation. Also, having accurate data can and will have a direct effect on costs.
COPYRIGHT 1990 Lippincott & Peto, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1990, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:testing
Author:Menough, Jon
Publication:Rubber World
Article Type:column
Date:Sep 1, 1990
Previous Article:Evaluating truck tire patents.
Next Article:Chemical degradation of polyurethane.

Related Articles
Spreadsheets in the rubber laboratory.
Sharpen spreadsheet skills.
Drug testing criteria should be narrower.
Technology Q&A: many ways to sum a column in Excel ... how to flip the order of data in an Excel table ... get Word to print single-character...
In Excel, cell names spell speed, safety: give a cell a name, and your work will go faster and be more error-free.
Ferret out spreadsheet errors: use Excel's tools to uncover and correct formula problems.
No distortion left behind: the New York Times education columnist gets it wrong.
What to do? A major dilemma has your brain cells spinning? Decisions, decisions ...

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters