The Hidden Killer:
How to Protect Your Designs

by Dr. Nick Henwood, Rotomotive Limited

The subject of creep, as it applies to rotomolded polyethylene articles, is an important subject, but one that is poorly understood. Frequently creep is not considered in the design of a rotomolded product, even when it should be. There are a number of reasons why this can be the case, the most important being the absence of readily available data and a general lack of recognition of creep as a factor within rotomolding.

This article seeks to promote a consideration of the subject of creep in the minds of designers of rotomolded products and to provide some signposts to a sensible approach in assessing whether creep is something that needs to be considered.

What is creep and why does it matter?

The term “creep” is used to describe the gradual deformation that occurs when a material is subjected to prolonged loading. In many plastics, especially semi-crystalline materials like polyethylene (PE), this continued deformation can be considerable over the period of an extended product lifetime. Significant deformations will occur even when the material is subjected to quite low stresses, well below the levels likely to cause yielding in a conventional tensile test.

Many rotomolded products are subjected to prolonged continuous loading and most of these products are expected to perform well for several decades of usage. For example, an underground tank will be subjected to continuous ground pressure and will be expected to be “buried and forgotten”. The influence of long-term creep effects may cause a product failure, years after the product has been put into service.

Unfortunately polyethylene (PE), the main material used to make rotomolded products, has very poor resistance to creep, even in relation to other commodity thermoplastics such as polypropylene and poly (vinyl chloride).

The design of critical rotomolded products will usually be carefully considered, but often only short-term material properties will be used to evaluate performance. A standard finite element analysis (FEA) will call for material properties such as Young’s Modulus, Yield Stress and Poisson’s Ratio; all of these are strictly short-term properties, measured in a matter of minutes of testing.

Requests for creep data from material manufacturers can sometimes be met with a degree of reluctance to share data openly. This is probably because evaluation of a long-term property such as creep will involve an extensive programme of test work. In some cases, manufacturers may simply not have reliable data on their grades. In other cases, they may consider that the costs they have incurred during data gathering are too high to enable them to share the data openly. Either way, this sounds like bad news for the rotomoulder!


What causes creep?

To understand the phenomenon of creep in semi-crystalline plastics like PE, it is necessary to consider the molecular structure of the material.

PE molecules are long chains of carbon and hydrogen atoms. In the molten state, all the chains take up a disordered arrangement, called the amorphous phase. As PE cools, some of the molecules start to fold along their length into regular structures, known as the crystalline state. For the PE grades typically used in rotomolding, approximately half the molecules end up within the crystalline phase and half the molecules within the amorphous phase. This is illustrated in Figure 1.

When solid PE is subjected to stress, the first deflections that are observed are due to an immediate distortion within the crystalline phase; this has been likened to the strain observed when a metallic spring is loaded with a weight. As such, the immediate deflection of the crystalline phase is reversible, if the load is removed.

Following distortion in the crystalline phase, a reaction of the amorphous phase can be detected. This effect is due to the molecules in the amorphous phase sliding over each other and re-arranging themselves as the stress continues to be applied. This distortion has been likened to the strain observed when a dashpot containing a viscous fluid is loaded with a weight; it is time dependant and non-reversible, even if the stress is removed.

These “spring and dashpot” models are useful to an extent, to aid visualisation, but the movement within a real material will be more complex and will, of course, be happening at a molecular scale. The distortion of the crystalline and amorphous phases are interlinked and inter-dependant; one phase effectively supports the other as stress is applied. The combined performance of the material is often described as visco-elastic, to signify the different responses of the two phases to stress.

In the most simple visualisation of creep, a sample of PE under stress will exhibit an immediate deflection (due mainly to the crystalline phase), followed by a continuous deflection over time (due mainly to the amorphous phase). At some point, the re-arrangement of the molecules within the amorphous phase may be such that it can no longer protect the crystalline phase from further distortion and, at that point, the PE sample will exhibit signs of catastrophic failure.


How is creep performance measured?

Being a long-term phenomenon, creep cannot be assessed by short-term tests so, by definition, creep testing takes a long time. For PE, a relatively short creep test would be 1,000 hours, which is approximately 42 days. Test of longer duration than this may be required; some authorities recommend that testing up to 5,000 hours (208 days) is necessary, in order to enable product lifetime predictions to be made for several decades.

Creep testing equipment can be relatively simple; a static load is applied at the start of the test and the deflection of the sample is noted at time intervals. ASTM Standard Test Method D-2990 provides comprehensive details which can be used to establish a suitable protocol. Essential elements of a creep test include:

• Multiple test rigs. For a full evaluation of a single material, identical specimens will be tested at 3 or 4 different stress levels, with 4 or 3 repeats of each required. This means that a total of 12 specimens will be used.

• A loading system that applies a constant stress to a specimen. The applied stress can be created by loading the specimen in tensile, compressive or flexural mode.

• Accurate measurement of deflection. This will usually be by some form of dial gage.

• A temperature controlled environment, where a set test temperature can be maintained over the full duration of the test. For plastics, ASTM recommends Standard Room Temperature (23°C) as a starting point for most evaluations.

• Deflection measurements to be taken and recorded throughout the test. In the early stages of testing, the time interval between measurements will be measured in minutes, whereas later in the test this interval may be hundreds of hours.

The style of test rig used by Rotomotive is illustrated in
Figure 2. After some pilot trials on different loading systems, a flexural stress system was chosen, as it was found to be more practical and ergonomic when setting up multiple test rigs. The material sample spans 64mm between supports and the load is applied via a plunger in mid span. Deflection is measured by a dial gage mounted above the plunger that applies the load.

A walk-in temperature controlled chamber was built from 100mm thick insulation panels. Electric tubular heaters maintain the set temperature via a control box (see Figure 3) that monitors the average chamber temperature via twin thermocouples and switches the heaters on and off. An air circulation system runs continuously, to aid temperature distribution within the chamber.

Deflection values are read manually at specified intervals, which are based on the ASTM recommendations. Figure 4 shows the time intervals generally adopted. Deflection values (in mm) are converted to strain (by a simple beam bending formula) and plotted as they are gathered. In this way, any likely transition from one phase of creep to another can be identified and time intervals can be reduced, if the sample is exhibiting signs of imminent failure.






How is creep performance assessed and how can I protect my designs?

Figure 5 provides an illustrative graph of the entire progress of a creep test, from initial deflection to ultimate failure. A relatively high level of stress has been applied in this case (approx. 8.5 MPa – 1,200 psi), in order to show the full range of data in a manageable time. Sample strain has been plotted against linear time.

After an immediate displacement (creating a strain of approximately 0.02), the strain increases over time to approximately 0.06. During this period (the primary creep phase), the rate of increase in displacement is initially rapid and then starts to level off in a reverse exponential fashion. However, after a strain of approx. 0.06 is reached, the behaviour appears to enter a secondary phase, where the strain vs. time curve takes a more linear shape. When the strain reaches approximately 0.09, the rate of strain increases more rapidly (the tertiary creep phase) and the sample yields at a total strain of approximately 0.11.

One suggested approach to defining the workable “lifetime” of a product is to specify the time taken from initial loading to the onset of the secondary creep phase. This would provide a conservative estimate of lifetime, well before material failure occurs.

The performance of the sample during the primary creep phase, where strain vs. time appears to follow a reverse exponential, can conveniently be assessed by plotting strain vs. log time. This is shown, for the same set of data, in Figure 6. The data roughly lies along a straight line in this presentation and a linear correlation calculation will provide a formula for this line.

Provided that the material stays within the primary creep phase, the line formula can be used to extrapolate the data, in order to predict material behaviour at times beyond the duration of a test. Most authorities will only recommend a data extrapolation of two logarithmic decades, ie 100 times the test duration. This would mean that a 2,000 hour test could be extrapolated to 200,000 hours (approx. 22 years).

Samples subjected to lower levels of stress will behave in a similar fashion, but the timescale from initial deflection to the onset of secondary creep will be hugely different. This is illustrated in Figure 7, which shows the time to onset of secondary creep for stress levels from 5.5 to 7.0 MPa (800 to 1000 psi). This relatively small range of stress creates a massive difference in “lifetime”; between 500 hours and 200 years!  Clearly creep performance is highly dependent on the stress applied during the lifetime of a product.

The major prerequisite for applying this type of extrapolation approach is that we are certain that the material will remain in the primary creep phase and that it will not enter the more unpredictable secondary and tertiary creep phases. For the high stress level (eg 8.5 MPa) case, this occurred at a strain of approx. 0.06. It is probable that this critical strain level, before the primary-secondary transition takes place, will be higher than 0.06 for lower applied stress levels. The effect of the molecules in the amorphous phase of the material sliding over each other and re-arranging themselves is, to a large extent, time dependent. If a high stress is applied, the molecules will be required to rearrange themselves more quickly and there will be a tendency for them to break under the sudden strain of movement. This phenomenon is still under investigation, but the early results of studies at Rotomotive indicate that this assumption is correct.

One practical approach to lifetime calculation and product design would be as follows:

1. From calculations using the usual short-term data, nominate a maximum level of stress that will be allowed to occur, anywhere in the product, when it is under load.

2. Conduct creep testing on the chosen material at this maximum stress plus a significant factor of safety (eg 50%), to establish at what strain the transition from primary to secondary creep occurs; this is the “critical design strain”.

3. Conduct creep testing at more realistic (i.e. lower) stresses and acquire linear regressions of the creep curves (strain vs. log time) so generated.

4. Define material “lifetime” at a given stress level as the calculated time to reach the “critical design strain”. This can be calculated by extrapolation of the strain vs. log time curve.

5. Decide what stress level provides a satisfactory material “lifetime” for the product being designed.

6. Adjust wall thicknesses and other features so that this stress level is not exceeded anywhere around the product, under all realistic design scenarios.

7. This approach will provide a conservative design, because the “critical design strain”, was determined at an unrealistically high applied stress.

This article has described the initial results of what, for Rotomotive, is a long-term study. Investigations will continue, in order to better define the effects of the various parameters on the creep performance of polyethylene roto grades.

Note from the Author:

This article has been written in the hope of stimulating more interest in the subject. Since creep testing is so long-term, progress is necessarily rather slow. With more test rigs, we could produce useful data at a faster rate and we would be very willing to discuss a cooperative approach with any interested parties. Our aim is to shed light on this important subject, for the benefit of the worldwide roto industry.