Picking the right plastic for a new product really is like looking for the proverbial needle in a haystack. There are at least 50,000 grades of commercial plastics on the market right now. And even a single materials family can consist of hundreds of subtly different grades from many different vendors. Nowadays, most engineers use the web to sift through this haystack. With just a few clicks, they can access data sheets from one or more suppliers. Then, after a cursory comparison of mechanical and physical properties, they pick a few likely candidates. And how close does this process get them to the right plastic for their application? Not very close at all, it turns out.
As a visco-elastic material, plastics have performance characteristics that depend not just on straightforward stress and strain conditions but also on strain rate, temperature, and time. "A lot of engineers forget about these factors, especially the rate dependence," notes Jim Lorenzo, an applications engineer with Bayer Polymers (Pittsburgh). Thermoplastic performance also depends heavily on environmental conditions. And it depends on part design and molding conditions too. All this dependence makes it notoriously difficult to translate application requirements into the mechanical and physical properties needed to pick the right plastic for the job. "They don't teach you how to do that in school," says Ranganath Shastri, a global knowledge management steward for Dow Plastics (Midland, MI) and a materials scientist who has spent much of his career debunking plastic property misconceptions. Wim Bruijs, head of global design engineering for GE Plastics, agrees. "Customers often ask 'What is the maximum stress allowed for this material?'" he says. "It's a fair question but tough to answer. Is it a continuous load? What's the temperature? What's the strain rate? The question probably indicates that they are not thinking deeply enough."
Faced with the complexity of plastics, engineers can make the mistake of oversimplifying things. In fact, plastic design experts working for some of world's biggest material suppliers report that even experienced engineers often put too much credence in the kinds of simple, single-point property descriptions found on data sheets. In doing so, they flaunt the warnings and disclaimers published right there on the data sheets themselves. "We see it all the time," says Shaun Tierney, an applications engineer with GE Plastics (Pittsfield, MA). "People take these properties literally when they shouldn't." And in his 25 years of plastics engineering experience, he has seen "countless examples" of materials selected on the basis of a single-point property—even though materials selected this way can result in part failures when subjected to the real-world conditions found a bit further out on that material's stress-strain curve.
But parts don't have to fail for plastic property misunderstandings to cause problems. Early in the design process, the misunderstandings have more subtle consequences, materials suppliers say. "It's easy to deceive yourself about the properties and rule out the best material candidates for all the wrong reasons," says Mark Matsco, manager of Bayer Polymer's applications engineering group. This particular deception can lead to wasted time and money as less-than-ideal plastics make it too far down the design path. Over-engineering also results, since some property misunderstandings understate stiffness and strength.
Drop the Data Sheet
So one of the most basic and important tips from the plastics pros has to do with putting data sheets in the proper perspective. Andrew Poslinski, of GE Plastics' global application technology group, likens data sheets to a resume for the product and describes them as useful only for screening and comparing materials—and maybe for plugging the values into simple, back-of-the-envelope calculations. "Data sheet properties can give you a quick feel for the material's ability to meet deflection, impact, and processing requirements," he says. "But only in a comparative sense." Problems arise when engineers read too much into data sheet properties and try to use them later on in the design process. "Can I design to these properties? No," Poslinski says emphatically.
And why not? Well, a common quip in the applications engineering departments at the material suppliers goes something like this: "Data sheet properties are an excellent indicator of final properties—if your product is a tensile bar." The standardized tests used to measure data sheets properties contain data measured in a lab under ideal conditions on test specimens that bear little resemblence to the geometry of real-world parts. These tests likewise take place at temperatures, strains, and strain rates that rarely correspond with real world conditions. A single number representing one point on a stress-strain curve, for example, can't begin to convey a plastic's behavior over a range of conditions. "The point is that this data is just a single point," says Poslinski.
There are other reasons not to take the data sheet properties as gospel. For one, they represent averages from several tests, rather than an absolute value. "Only the average value is given on a data sheet. You never know what the spread is," Shastri notes. For another, data sheet properties don't capture that crucial link with manufacturing. "The injection molding process influences the final mechanical properties of the plastics," Tierney says, citing the relationship between molding conditions, internal stresses, and final part strength as just one example.
Plastics experts say the best bet when screening materials is to start with multi-point data and plenty of it. "Don't just ask for a stress strain curve. Ask for several of them at the temperatures, strains, and rates that correspond to your application," Shastri says.
Ask and You Just Might Receive
One reason more engineers don't ask for enough multi-point information upfront and continue to rely on data sheets clearly comes down to availability. Data sheets are always just a few clicks away. "But it's a mistake to think they're all that's available," GE's Tierney says. In fact, a bit of digging can yield lots more information, and some of that data is easily available on the web. The CAMPUS database, for example, contains standardized materials data from dozens of suppliers, and plenty of that data is multipoint, Shastri says. For example, CAMPUS presents creep data for many materials in tension in at least three temperatures and four stress levels. A proprietary database from IDES (Laramie, WY) also serves as a source for single-point and multipoint data from many suppliers. Material suppliers themselves make some multipoint data available on their websites.
If you still don't see what you need, ask. Materials suppliers have developed a lot of data on impact, creep, and allowable design values that they share only with customers or potential customers. "The expectation of design engineer is that the supplier will have properties on every material to 200C in one-degree increments at a variety of strain rates. That's pretty unrealistic," Shastri says. But he adds that a good materials supplier can usually suggest a suitable equivalent grade for which data does exist. And for the right incentive, like selling a lot of plastics, they just might generate it for a customer.
Obtaining additional property data early in the design process won't address all of the more complex issues that arise as a design matures. Thermal problems, high-rate impact, fatigue, and NVH, to take just a few examples, often require advanced computer simulations and extensive testing. So do applications that will subject plastic parts to chemicals, UV light, and other degrading environmental influences. Yet materials experts all argue that some extra attention to five basic—and commonly misunderstood—properties can head off many design mistakes and even pave the way for the more advanced work as the design cycle unfolds. Here's a closer look at those five:
To gauge the impact resistance of a plastic, many engineers start with the Notched Izod number that's right under their nose on just about every data sheet. But like most data sheet properties, this common test, which involves dropping a weight onto a notched sample, doesn't correspond to end-use environments. Tierney describes it as "a number no one knows how to use." The number does have some value as an upfront comparison and indication of a material's notch sensitivity. But it doesn't say much about real-world impact behavior, which hinges on part geometry, stress states, strain, rates, and temperature, Bruijs points out. "Impact is a very involved area," he says. Bayer's Lorenzo agrees and adds that engineers who otherwise have a handle on impact tend to miss the rate sensitivity in particular. "I'll hear comments like, 'My part only has a 5-mph impact, so I can get away with quasi-static data from lab tests,'" Lorenzo says. But the impact performance suggested by these two types of data can be off by "orders of magnitude," he says. In Lorenzo's view, it's equally important to recognize that strain rate, unlike initial impact speed, can vary locally within a plastic part. "Engineers need to be aware of this fact and determine the actual strain rates related to their application," he says.
Despite the complexity of impact, materials suppliers say some engineers still try to specify materials with just a single data sheet property in mind. "We get a lot of calls from people saying 'The Notched Izod is poor here, do you have something better?'" Bayer's Matsco says. "And they don't need it in the first place." What they usually do need, he and other materials experts say, is instrumented impact data from tests such as the Dynatup impact test. This type of test measures deflection associated with a range of impact loads. "It provides more useful information such as energy at a given strain or total energy at break," Shastri says, noting that this kind of multipoint data can pave the way for computer simulations as the design process unfolds. Unfortunately, instrumented impact data is harder to come by than the Notched Izod value pulled from a data sheet. But all the suppliers here reported that they do have at least some multipoint impact data available for many of their materials.
With stiffness representing such a fundamental engineering concern, it's not surprising that flexural and other modulus values give many engineers headaches. "Unfortunately, there are too many different moduli to choose from," says Mark Minnichelli, director of commercial technology for BASF Corp. (Mount Olive, NJ). Engineers typically start with flexural modulus since it's the most readily available indication of a plastic's rigidity.
Or is it? As Bayer's Lorenzo explains, the test for flexural modulus takes place under such low strain conditions, well under 1 percent, that it refers only to the linear elastic portion of the stress-strain curve. "It's done on tensile bars that bear little relationship to end-use geometry," he says, noting that final stiffness of a part depends very much on its thickness, which will usually differ from that of the tensile bar. For this reason, Lorenzo recommends that flexural modulus be used only for screening purposes or in simple stiffness calculations. "But only to 1 percent strain," he says. "Anything beyond that and you would tend to over-predict stiffness."
Some engineers also use flex modulus as a rough stand-in for tensile modulus below 1 percent strain. And on many stress strain curves, they do look pretty close. Dow's Shastri says to be careful though, because flex modulus will in fact be higher. "If you plot flex versus tensile, you'll get a ratio of 1.3 to 1.6," he says. Using flex modulus would again tend to overpredict the stiffness of the final part. For these reasons, Shastri takes a dim view of flexural modulus and would like to ban it from data sheets altogether. "It's not a very good representation of material behavior, but we can't get rid of it," he says. "Our customers still ask for it."
Shastri suggests engineers stick with tensile modulus for most initial screening purposes because of two important factors. First, tensile modulus data doesn't depend on part thickness as flexural modulus does. "Tensile modulus calculations assume a uniform stress distribution through the part's cross section," he says. "With flexural testing, there's a stress gradient. So you have to hope stress passes through the part evenly from top to bottom, even though it may not." Second, tensile modulus, when used during initial screening, can be carried through to simulation later in the design process. "CAE codes need a tensile modulus, not flex," Shastri points out.
Yet tensile-modulus has its problems too. Shastri notes that the test samples are pulled at a much slower rate and to much lower strains than real-world applications usually see. In the ISO procedure, for example, tensile modulus testing takes place at 1 mm/min and at 0.005 to 0.025 percent strain. "In real life, a plastic part might see 20 times that rate," he says, adding that the faster rate corresponds to a higher modulus. Tensile modulus can thus cause engineers to underestimate real-world stiffness and "end up with a thicker part than they really need," he says.
Finally, experts caution engineers not to forget about secant modulus, which comes into play when an application's strain levels go beyond the non-linear portion of a plastic's stress strain curve. "Engineers tend to look at modulus values only at the lowest strains," says Mark Schuchardt, a design leader for DuPont (Troy, MI). That strategy will work for many plastics. "A glass-filled nylon may not have to endure more than 1 percent strain," he says. But he adds that some plastics, like elastomers, might routinely be subjected to high strains that would require the use of secant modulus. And he cautions engineers to look closely at the secant modulus value. "The value corresponding to the actual strain in your application may be entirely different than the book value," he says.
Creep, or the non-reversible deformation of material under a load over time, represents another frequently overlooked or misunderstood plastics property. Some materials suppliers attribute this blind spot to a longer engineering history with metals. "Unlike metals, plastics do creep at room temperature," says Lorenzo. Even when design engineers do recognize that fact, creep can confound their initial screening efforts because "data sheets provide no direct indication of creep," Poslinski says.
Some engineers try to infer a material's likelihood to creep by looking at various data sheet temperature values—like heat deformation temperature (HDT), glass-transition temperature, or Vicat softening temperature—reasoning that temperature performance might give a rough indication of a material's creep resistance in some comparative sense. According to DuPont's Schuchardt, these single-point temperature values can indeed provide some rough comparative insight. But he quickly adds that so many factors affect creep—including fillers and reinforcements—that rules of thumb because meaningless beyond the screening stage. "Every material is a different story," he says. GE's Poslinski agrees. "Really understanding creep is more involved and requires multipoint data covering stress state, temperature, and time," he says.
Dow's Shastri says he routinely sees engineers ignore creep when they anticipate only small loads on the plastic. "That's a big mistake," he says. As an example, he mentions television cabinets. "Even a light weight item placed on top, like a set top box, could cause the cabinet to creep," he says.
Creep difficulties often pop up when design engineers try to determine part thickness using a single-point, room-temperature modulus value from a data sheet or online source rather than creep modulus—or the apparent modulus of the material under load over a long time period. "You tend to make the part thinner when you ignore creep," Shastri says, explaining that the creep modulus can be substantially lower than the modulus listed on the data sheets. So why don't more parts fail? Shastri believes that the typical design engineer, through the practice of applying a safety factor to modulus, unwittingly accounts for creep. "Safety factors helps indirectly," he says.
Creep modulus, however, is no cure-all. Lorenzo reports a common mistake related to its application. "The danger of creep modulus is that it's way over-conservative," he says. In his view, it tends to overpredict deflection, making parts "look weaker than they really are." This prediction error occurs because engineers tend to apply creep modulus to an entire part, assuming that the whole thing will undergo the same stresses over time. In reality, only the high-stress regions of the part, those most subject to creep, may need to have that lower creep modulus applied to them. "You end up looking for materials with a much higher initial modulus than you really need, which may turn you away from plastics as a solution," he says. Or it can result in design errors, like overly thick parts. As a solution to the creep modulus problem, Lorenzo relies on computer simulations. He explains that the visco-elastic material models in commercial CAE codes in essence apply a different creep modulus to each individual stress zone. Modeling provides an opportunity to account for creep through design rather than looking for materials with an unrealistically high modulus.
And Shastri point out that modulus misunderstandings carry over to creep as well. For the same reasons that he prefers tensile to flexural modulus, he believes that creep modulus under tension is more useful than creep in bending, though the latter is still easier to find.
Over the years, HDT has become shorthand for whether a material can take the heat. "Some people look at the HDT value and think that's the temperature their part can survive under or not be able to exceed. That's just wrong," Minnichelli says. "The misuse of HDT has been a pet peeve of mine for 15 years." He adds that he would like to leave it off data sheets altogether. So would Shastri, who call it "completely arbitrary test." And with many engineering materials, the HDT doesn't get anywhere close to a real design value. "With crystalline materials, it's certainly not within the range you would design to," says DuPont's Schucardt.
HDT's chief shortcoming, like those of most other data sheet properties, comes down to the disjunction between arbitrary test conditions and real-world parts. The two HDT tests the load test specimens to 66 and 264 psi, respectively, and then record the temperature at which a given deflection occurs. According to Shastri, lore has it that the Dow scientist who came up with the HDT test used 66 and 264 psi because those loads corresponded to some metal weights he had on hand at the time. "One disc corresponded to 66 psi and he had four discs in all," he says.
As an alternative to HDT for material screening, Shastri prefers Vicat softening temperature, which measures the temperature at which the surface of a sample starts to soften. "It's pretty close to most plastics' upper use temperature in the absence of a load," he says. An even more useful indication of temperature performance comes from plots of modulus over a temperature range. They give an obvious graphic indication of where temperature effects become pronounced, Shastri says. And as the design matures, our plastics experts unanimously advise engineers to obtain stress-strain curves to cover the range of temperatures expected for the applications.
With the upswing in the use of glass-reinforced materials, engineers should give some extra thought to the tensile strength and stiffness information culled from data sheets and on-line sources. BASF's Minnichelli points out that part and tool design, as well as molding conditions, ultimately dictate the orientation of fibers in real parts. "Engineers have to have an awareness that glass fibers don't uniformly orient and what that means for stiffness, strength, and shrinkage," he says. "Don't assume that the fiber direction for a real part is the same as the fiber direction in the test specimen." To account for the anisotropy of glass-reinforced parts, some suppliers do provide strength data for both the flow and transverse directions. The former data comes from standard tensile bars, with most of the reinforcing fibers lined up with the bar's long axis. The transverse-direction properties, by contrast, are measured on samples cut from a larger part with the fibers oriented largely perpendicular to the direction of flow The strength and stiffness properties in this direction can be substantially lower, materials suppliers report.
This kind of data certainly helps, but it too should be taken with a grain of salt. Engineers can certainly try to apply the lower transverse-direction properties to those areas of the part where the expected fiber orientation isn't in the flow direction. That approach only goes so far, though, because fibers in real parts don't orient themselves in 90-degree increments and in just two dimensions. "Transverse and flow direction numbers don't take into account the orientation in the third direction," Shastri notes.
And while fiber orientation affects the entire part, it really matters at the "knit lines" formed wherever flow fronts come together. Reinforcing fibers don't span knit lines, so they create an inherent weak spot in the part. Minnichelli says he sees knit-line failure "all the time," especially in impact situations. "Unfortunately these problems tend to crop up further down the design path," he says. For this reason, and because of the need to predict fiber orientations, Minnichelli strongly recommends mold-filling simulations when dealing with glass-filled parts. "It's necessary to go to sophisticated modeling tools when dealing with fibers," he says.