Self-Sensing Composites



Introduction


The concept of integrating sensing into a material was birthed from the fields of biomimetics, multifunctional materials and structural health monitoring. Biomimetics is "the application of methods and systems found in nature to the study and design of engineering systems and modern technology." [1] Meanwhile, the field of multifunctional materials seeks to give an engineering material additional functionalities; examples include structural materials with energy harvesting , or permittivity tuning capabilities. Finally, structural health monitoring looks monitor the health of a structure or vehicle through the use of not only sensors, but also computation and analysis.


In designing composites with integrated sensing we look to incorporate both the sensors and the computation inside the composite in a distributed addressable network analogous to that which can be found in an animal nervous system. This demands investigation into several fields of engineering: mechanical, electrical and computational.



Mechanical Investigations


Adding sensing to a composite material such that the end product acts as a material with a nervous system inherently involves placing a "nervous system" inside the host composit. In this case our "nervous system" is - in fact - sensors and microcontrollers which will act as inclusions and stress concentrators in the host composite. Damage induced by the peak values of stress concentration around the embedded devices is, in fact, one of the main concerns. To assess this and related issues, we have fabricated a series of samples with and without embedded (dummy) sensors and micro-processors in S2 glass fiber/epoxy laminates, and systematically tested the samples while continuously monitoring the response by the acoustic emission technique (for tensile tests only). In this manner we have sought to address the process of damage initiation and evolution within the material. Ideally, embedding geometries and practices can be optimized to limit the detrimental impact on the composites structural properties.


For the purpose of our research we are using a unidirectional S2 glass - epoxy resin matrix prepreg for our host material. The reasons for the choice as opposed to a carbon fiber composite stem from cost and processing considerations and the fact the failure modes and acoustic wave propagation are similar enough between the two for corresponding geometries and loading conditions.


Short beam shear (SBS): Short beam shear tests are performed in accordance with ASTM D 2344 standards and tested in a MTS model 309.20 load frame with 22.2 kN load cell.


Quasi-static SBS strength: Tests are performed in displacement control with crosshead rate of 1.25 mm/min with failure consisting of a 30% load drop-off, two-piece specimen failure or excessive travel. Six specimen of each configuration (shown below) were used.


SBS Fatigue: Fatigue was performed using a 10 Hz sinusoidal input function with a stress ratio of R = 0.1. Failure determined by excessive travel ( travel > 2.54 mm).


Sensor embedding options


Fatigue short beam shear tests indicate that the host composite material is very sensitive to the local stress concentration introduced by the embedment as it results in a decreased fatigue life. Additionally, it has been found that the geometries involved have a profound effect on the fatigue life of our host composite.


Fatigue results for various geometrical inclusion considerations.


Tensile Tests: Tensile tests were performed under quasi-static and fatigue characterization using MTS load frame with hydraulic grips on dog-bone shaped samples. An acoustic emission system - Physical Acoustic Corporation's PCI-2, R50 and micro30 sensors - is used to monitor matrix cracking, delaminations and fiber breakage events and their locations. The material properties were assessed testing unidirectional [0]n, [90]n, [±45]n laminates instrumented with strain gages to measure the transversal and longitudinal strains. For a better understanding of the acoustic emission results, as well as for assessing the early damage initiation in laminates with integrated dummy sensors we dedicated part of our efforts in exploring the material microstructure. Standard optical metallographic techniques allowed the characterization of void size, percentage and distribution of flaws as well as the analysis of the sensor-matrix resin interface.


   

(Left) Typical failure of [45]5s samples with integrated dummy sensors.
(Right) Micro-crack initiation at sensor and resin pocket eyelet in [0] to 5% strain.


The outcome of this research highlights a significant difference of the mechanical behavior, damage growth and location in samples with integrated implants and different lay ups. Overall, the data acquired by testing different laminates is consistent and indicates similar trends in all the experiments. Some of the major results follow.


The failure mechanism initiates at the dummy sensor location in [0]n , [90]n and [45]n laminates. Stress concentration due to the presence of material and geometrical discontinuities is certainly responsible for early micro-cracks events around the embedded device. In [0]n the micro-damage propagates freely along the fiber orientation within the resin matrix. Nevertheless, the material tensile strength seems not to be compromised.


A net cross section failure at the sensor location was instead observed in [90]n laminates with a significant reduction of the material overall strength (24%). Finally, [45]n laminates with embedded chip resistors showed a different failure behavior. Although subjected to shear loading and in presence of stress concentration areas at the dummy sensor-resin matrix interfaces, the failure of the sample never occurs at the sensor location.


The aforementioned conclusions demonstrate that non-zero fiber orientations seem more suitable for the embedment of sensors and devices within glass/fiber laminates made by prepreg material, although the laminate lay up optimization for bearing the desired loads is, of course, needed. Furthermore, the damage initiation detected in samples with embedded devices has been always identified at the sensor location and characterized by high values of the signals amplitude so that the enhancement of the embedment and manufacture techniques will be of paramount importance for the final material performance.



Computational Investigation


This aspect of our research can be clearly divided into three camps: mechanical modeling, network design & modeling, and algorithms. Mechanical modeling is necessary because without it we have no real understanding of the physical principals at play in our system and experiments. Network design and modeling must be considered carefully to ensure that the distributed network is robust, efficient and performs its functions as expected. Finally, algorithms are what takes sensor data and turns it into a quantification of the health of the composite material.


Mechanical Modeling: Embedding micro-sensors in composite laminates produces material discontinuity around the inclusions. This in turn produces stress concentrations at or near the inclusions. Both 2D plane strain and 3D FEM models have been developed to analyze the stress/strain state surrounding the embedded micro-sensors within a unidirectional composite laminate. The objective of the present numerical effort is to take into account the observed resin-rich areas caused by embedment, and to determine their effects on the local stress field around the embedment and the corresponding potential failure modes.


Finite element mesh of embedded sensor

Micrograph of a section of S-glass/epoxy composite laminate with embedded simulated micro-sensor with local finite element mesh of 2D FEM model overlayed.


Results: From the output strain field, it appears that the maximum longitudinal and shear strains occur at the sensor corners within the resin-coating interface (resin rich region). Among the three strain components, the shear strain has the maximum value. Since the neat epoxy resin can carry the least strain among all the material components in the composite, shear debonding at the resin-sensor coating interface is expected to be the main cause of failure initiation.


The three stress components along the resin-composite and resin-sensor coating interfaces generally have the same trend. However, in the latter case, the magnitude of stress is higher. The maximum longitudinal and transverse stresses occur at the sensor corner area. At the end of the resin pocket, the transverse stress increases. However, this small increase is not enough to cause fiber-matrix splitting. The shear stresses along both interfaces show sharp rises at the sensor corners, decreasing monotonically away from the singular point.


Based on the stress distribution results from, the values of the stress applied at the far end that cause the initiation of the failure were calculated both for the composite area and the neat resin area separately.


For the composite domain, maximum stress criterion is applied. Failure is expected to occur when at least one stress component along one of the principal material axes reach its corresponding strength. The criterion is defined as follows:


- Tension failure:


- Compression failure:


- Shear failure:


Here, the subscripts 1 and 2 refer to principal material axes of the composite material. For the neat resin area (including the transition resin layer and the resin pocket), von Mises criterion is applied. Failure is expected to occur when the stress components satisfy the chosen failure criterion. The von Mises criterion is defined as follows:


Under tensile loads, the initial failure is expected to be matrix cracking at the sensor corners in the resin-sensor coating interface at an early stage of the loading process.


Network Design & Modeling


Algorithms



Sensor & Signal Conditioning (Electrical) Investigation



Related Papers >

< Top of Page >