Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

Lightweight aircrafts represent the primary goal of aerospace industry in the close future. Being able to build up a CFRP (Carbon Fiber Reinforced Plastic) based aircraft will allow for a 15 % weight loss of the airframe that, in turn, will generate considerable savings in aircraft operation (fuel, CO2 emissions) and maintenance [1]. However, CFRP components are not riveted; their bonding is primarily based on adhesive technologies and, currently, no NDT technique has been developed and validated for ensuring the quality of the CFRP adhesive bonds. Low quality bonding would lead to potential cracks with imaginable consequences. In particular, surface contamination is a primary concern being cause of low bonding quality. Hydraulic fluids spills, moisture and release agents residuals represent the most common and dangerous contamination sources.

Operative needs for an NDT tool would require contamination detection to be performed within a few seconds in presence of interferents. Portability is also an important requirement for allowing easy operation in maintenance and repair shops. For this reasons, e-nose technology represent an interesting solution to develop an ad-hoc NDT tool for surface contamination detection in this scenario.

In this work we show the advancements in the screening process of artificial olfaction technologies for the above requirements.

Experimental

The e-nose measuring procedure is illustrated in Table 75.1.

Table 75.1  E-nose measurement cycle parameters. Response sampling rate was 4 Hz. Total measurement time was 25 s. Analysis of different specimen were carried out with 30 min delay
Full size table

The preliminary screening focused on the possibility to identify Skydrol contaminated CFRP samples from two uncontaminated control groups. A 50 % Skydrol/Water mixture was heated to 70 °C and stirred for 14 days. Oily and water phase were separated and CFRP samples immersed in the “Skydrol” water phase (pH 2). After 4 week, samples were removed and dried with a paper towel. Finally, they were covered with aluminum foil and stored at room temperature under N2 atmosphere with the aim to prevent any loss of contaminants from the CFRP samples.

For measuring purposes, samples were immersed in vials containing water and sonicated in order to extract Skydrol volatiles components to be presented to the polymer electronic nose sampling system. Each sample was measured five times and different samples were measured respecting a 30 min resting interval. A dataset of 15*5 samples was eventually collected to build an appropriate training set. A validation set of 15 samples was also collected 24 hrs later in order to evaluate the generalization and stability characteristics of the proposed approach. Training set LDA analysis shows a significant separation among the contaminated and control samples (see Figs. 75.1 and 75.2).

Fig. 75.1
figure 00751

Sensor array responses obtained by exposing the polymer sensor array to Skydrol/water mixture, pure water and diesel oil

Full size image
Fig. 75.2
figure 00752

3D LDA map obtained by 15 training samples. Separation between Skydrol contaminated and control groups is evident

Full size image

Results and Discussion

A polymer sensor based commercial solution has been first used for demonstrating the capability to detect SKYDROL500-B (a highly aggressive hydraulic fluid) contamination analyzing a Skydrol/water mixture at 0.6 % v/v [2, 3]. The response of the e-nose was recorded (see Fig. 75.3) and characteristic response features including max response, max climb rate, max descent rate and AUC, were extracted. An LDA analysis demonstrated the discrimination capability of the polymer array correctly separating Skydrol samples from Diesel Oil samples and control (water) samples.

Fig. 75.3
figure 00753

3D LDA map obtained by 15 validation samples. Separation between Skydrol contaminated and control groups is consistent with training set map

Full size image

This system was able to correctly classify and recognize 100 % of contaminated samples (Table 75.2) showing the feasibility of the electronic nose approach to the Skydrol surface contamination detection. However compliance with operative requirements is still to be reached. Furthermore, polymer sensors have shown to be prone to significant drift effects. This, in turn, would require frequent recalibrations of any sensor array relying on them. Eventualy, this would complicate the operation of the e-nose. Future analysis will be actually performed with an hybrid e-nose architecture that will allow to test the use of more stable electrochemical, Metal Oxide and IMS technology based sensor array. The use of an integrated IR surface heater has also been devised to increase the volatile uptake by the e-nose sampling system.

Table 75.2 Contaminated CFRP samples confusion matrix obtained on 15 validation samples, classified by an LDA classifier trained on the training set. A single misclassification error occurred among the two different control classes
Full size table

Conclusions

In this work, preliminary steps for the screening process of artificial olfaction technologies for the CFRP surface chemical contamination are reported. A polymer sensor based commercial e-nose platform was successfully employed in the discrimination and detection of hydraulic fluid contamination over CFRP surfaces. However compliance with operative requirements is still to be reached.