Defect Detection and Classification in Aerospace Materials Using Phased Array Ultrasonics

The use of phased-array ultrasonic instruments and probes allows for optimization of both the sensitivity and resolution for detection of different defect types through electronic beamforming, scanning, and focusing processes. However, development of these inspection processes requires an extensive database of material specimens containing known defects. Through the collaboration of the aerospace NDE community, the University of Dayton Research Institute (UDRI) has begun development of an ultrasonic database of inspection results for a variety of engine alloys and defect types. Inspections have been conducted on specimens containing a variety of synthetic and naturally occurring inclusions, voids, and defects, using the Turbine Engine Sustainment Initiative (TESI) phased array ultrasonic system. This paper will discuss current progress in the development of the ultrasonic database, and methodologies for embedded defect detection criteria. Introduction: The use of ultrasonic inspections for the detection of hidden defects has been used extensively throughout the aerospace industry for structures, pre-production and serviced engine components, as well as material processing quality control [1,2]. This vast range of inspection applications leads to a need for inspection systems that are versatile, but can also can be customized for the detection of whatever target type is required within the parent material and component geometry of interest. The inspection applications become particularly challenging for serviced engine components. Although these components have been inspected multiple times during the pre-production stage for processing defects, the undetected defects can become altered during subsequent forming operations. Voids and inclusions, which were below the detection threshold for new components, may develop new features during service. Changes in material microstructure induced by stress and temperature result in void growth or coalescence, which could lead to crack incubation. These changes in the defect morphology may require reinspection of the components throughout the life of the engine. The challenge to the inspection developer is to identify the pass/fail criteria for components that exhibit critical defects, while minimizing the likelihood of condemning serviceable components. Development of a truly reliable inspection is extremely difficult in the absence of specimens that contain actual defects to be detected. In that case, inspection requirements are often developed based on practical experience with the ultrasonic response of the material and defect. This methodology for developing inspections makes use of inspector experience, which is highly subjective, and not easily quantified for the purpose of obtaining probability of detection (POD) curves. For some inspection systems, reliability is based on the probability of detecting a machined target, such as flat bottom holes (FBH) or side drilled holes (SDH) of various sizes. Correlations between the response from machined targets upon which the POD curves are based, and the natural targets to be detected are often limited however, and thus the probability of detecting the natural targets is largely unknown. Significant improvements in the reliability of the POD curves would be expected if synthetically produced targets, which represent the type and range of reflectivities expected from the natural defects were used. Specimens produced with targets that are more representative of the actual inspection condition could then be used to generate POD curves that are directly related to the probability of detecting the defects of interest. Specimens that contain synthetically produced natural targets over a range of reflectivities, can also be used to develop defect detection criteria. Analysis of the A-scan waveforms may provide information about the target that can be used to identify the defect type. Previously, collecting A-scans during an inspection was impractical due to the memory required for data storage. Current technology however, permits the acquisition and storage of large quantities of A-scan data, thus permitting a new approach to defect detection and classification. Current inspection procedures make use of C-scan amplitude data only to make pass/fail decisions. These decisions are derived using various algorithms based on a calculated amplitude within the suspect region, which is compared to a calculated amplitude of the surrounding material background noise level. These algorithms are very carefully controlled and monitored since they are what ultimately decide the risk level of the inspection. Since aerospace components can contain a variety of defects however, it would be beneficial to determine not just the relative “size” of the target, or it’s severity based on a signal to noise ratio, but the actual type of defect that has been located based on some feature or property of the A-scan. Under the Turbine Engine Sustainment Initiative (TESI) a new approach for determining POD and defect detection is currently being developed. Specimens are being produced that will contain selected targets representing the range of reflectivities expected for embedded defects in turbine engine components. These targets currently include carbon, alumina and tungsten, with calculated reflectivities of 0.8, 0.34 and 0.08 respectively, in Rene [3]. These specimens (Number 1 and 4 in Table 1) were produced with a titanium or Rene matrix material, which was ultrasonically clean with very little background material noise. The embedded defect targets examined in the current study are 3mm spheres. Good mechanical bonding between the target and matrix material was insured through ultrasonic and metallographic examination. In addition to the spherical targets, 2 target locations were prepared but did not contain spherical targets during final specimen processing. These locations remained as a discontinuity in the matrix material after processing; however the exact morphology of the remaining void is unknown. The simulated void targets are included with the spherical targets in the specimen description as an example of an irregular void type defect. Based on the ultrasonic response from these targets, advanced methods for defect classification, based on A-scan correlation and classification methods, will be explored. In addition, the ultrasonic response from these new specimens will be compared to the ultrasonic response from traditional machined and synthetically produced targets. Through the cooperation of the NDE aerospace community, many specimens have been made available to the TESI program for this comparative study. During the study, inspections will be conducted on all available embedded defect specimens, and the data compiled to create an ultrasonic database. The database will include data from a wide variety of targets, transducers and inspection conditions, which will be made available to the aerospace NDE community. Ultimately, this database could be used to assess the detection capability of new systems and techniques as they become available. A summary of the embedded defect specimens that have been made available to the TESI program thus far is shown in Table 1. Inspections on these specimens are currently in progress, and will continue as new specimens and inspection techniques become available. The discussion in this paper will focus on the ultrasonic response from targets roughly 12mm deep, within a nickelbase superalloy matrix. (Specimens #4-8 in Table 1). Results: The experimental setup used in this study consists of the R/D Tech Tomoscan III pulser/receiver, gantry x-y scanning system, conventional single element and multi-element linear array transducers. The data presented in this paper were acquired using 10MHz, 3” focus, 0.375” diameter single element transducer, (UTX, Inc. 112 Milltown Road, Holmes, New York) and a 10MHz, 64 element, (0.45 X 5.0mm element size) linear array transducer (R/D Tech Ultrasound Transducers, 60 Decibel Rd., State College PA). Although linear array transducers have been widely used in applications such as pipeline, power generation and weld inspections [4,5], the use of phased arrays in aerospace ultrasonic inspections is not as common. Therefore, it is of value to compare inspection results obtained using both conventional and linear array transducers on aerospace relevant materials and targets. This paper will discuss the results obtained thus far in the development of inspection processes for turbine engine components using phased array transducers. A more detailed comparison of the ultrasonic response from phased array and conventional transducers is discussed elsewhere in these proceedings. Table 1. Embedded defect targets examined under the TESI program (current status).