Log in

Investigating CFRP Material Property Uncertainty with OnScale

By Chloe Allison 10 February 2020

A benefit of OnScale simulation is the ability to quickly perform sensitivity studies by running hundreds of simulations in parallel for varying input parameters. An example of this is a material property sensitivity study that investigates how varying material properties affect a simulation output. This is hugely beneficial as the material properties implemented in a simulation can often be the largest source of uncertainty. This is very often the case when simulating carbon fiber reinforced polymer components (CFRP): material properties are often implemented based on experimental measurements. However, the accuracy of these experiments is often unknown and the material properties can vary between different material batches and specimens. Through the use of a sensitivity study in OnScale, these material variations can be quickly investigated to analyze the impact of material properties on the inspection.

A CFRP laminate is constructed from multiple layers of carbon fibers that can have varying orientations as shown in Figure 1 (left). When performing pulse-echo inspections of these structures, it is common to observe reflected signals from the internal ply layer structure, as shown in Figure 1 (right). In this received signal we can observe a large frontwall signal followed by the internal structure reflections, which decays in amplitude with increasing depth into the sample, until we see the reflected signal from the backwall. The internal reflections are generated from resin-rich layers between the composite ply layers and can create challenges to identify defects within the structure. However, these signals can also provide a benefit with the ability to characterize the internal structure and identify any out-of-plane waviness.


Figure 1. (left) Section of a microscopic image of a CFRP laminate showing (top to bottom) 0°, (±)45°, 90° and  (±)45° ply layer orientations. (right) OnScale simulated A-scan signal from an ultrasonic inspection of a CFRP laminate displaying the frontwall reflection at 3μs, backwall reflection at 7.5 μs and multiple internal structure reflections in between.


A material property sensitivity study was conducted in OnScale to investigate how variance in material properties affects the reflected signal from the internal CFRP structure. Two input parameters, the out-of-plane modulus (E22) of the CFRP ply layers and the resin layer Young’s Modulus, were selected as input parameters. The nominal value for these material properties can be seen in Table 1 for an IM7/8225 CFRP component.


Table 1: Nominal material property values for IM7/8552 CFRP component


A sensitivity study consisting of 441 simulations was set up to vary both input parameters from their nominal value by ±10% in 1% increments. The reflected signal amplitude from the laminate resin layers was extracted from each simulation output. These values were then utilized to calculate the change in the signal amplitude with respect to the output from the nominal material property simulation. The variation in signal amplitude is shown in Figure 2 as a surface plot against the two input parameters.


Figure 2. OnScale-calculated backwall signal amplitude error due to variation in the out-of-plane elastic modulus of the CFRP plies and the Young’s modulus of the resin layer.

It can be seen from this result that the amplitude of the backwall reflection is sensitive to changes in the input material properties. The changes in material properties effect the transmission and reflection coefficients of the material interfaces. This results in variation in the backwall amplitude due to changes in the signal attenuation due to these internal reflections. This highlights the importance of selecting suitable simulation inputs to ensure derived results are representative of experimental performance. However, with the ability to run large parametric sweeps in OnScale, the impact of input parameters on outputs can be studied to gain insights to better understand experimental performance.

These simulations were all run in parallel with each individual simulation completing in 82 seconds on a 2-core cloud CPU with a memory requirement of only 30 MB. This meant the full sensitivity study data set containing 441 simulations was obtained in under 2 minutes. Without the ability to run all the simulations in parallel, this same study would have taken 10 hours to complete.

To learn more about this material property sensitivity study, download our newest whitepaper here!

Chloe Allison
Chloe Allison

Chloe Allison is an Application Engineer at OnScale. She received her MA in Electrical and Electronics Engineering from the University of Strathclyde. As part of our engineering team Chloe assists with developing applications, improving our existing software and providing technical support to our customers.