Table of Contents
1. Introduction
2. Flaw Detection
3. Flaw Characterization
4. Test Results
5. Factors Affecting System Design
6. Summary and Conclusions

1. Introduction

Advances in computer and digital technology over the past decade have made possible the development of high performance automated inspection systems for aluminum plate products. An ultrasonic scan inspection is normally performed by sweeping an ultrasonic transducer over the surface of the plate in an X-Y or raster pattern, alternately scanning the length or width of the plate in one direction and then indexing a fixed amount in the other direction. Data are recorded at points spaced a preset distance apart in the scan direction, and the spacing in the index direction is equal to the increment between successive scan passes.

The two primary goals of ultrasonic plate inspection are typically flaw detection and flaw characterization or classification, with porosity and inclusions being the most common anomalies. Cracks can also be readily identified through ultrasonic techniques. These two goals are sometimes achieved simultaneously if the ultrasonic data necessary for detecting flaws is also sufficient for characterizing them. Often, however, a fast and spatially coarse scan is performed to quickly identify areas containing flaws, and a higher resolution, slower speed re-scan is performed of just those areas in order to properly characterize them.

Aluminum plates lend themselves to this two-tiered inspection approach because in a typical plate, flaws can be detected at a coarser resolution than required for characterization, and the areas that need to be characterized normally constitute a very small percentage of the total (typically less than one percent). The total time to completely inspect a plate consists of the time for flaw detection plus the time for flaw characterization. If either is unnecessarily lengthy, it may become impossible or impractical to achieve 100 percent inspection.

2. Flaw Detection

Simple timing calculations illustrate why it is important to investigate techniques for making the detection inspection as fast as possible. For example, consider a plate with dimensions of 2.5 meters by 12.5 meters, 12.5 mm thick, that is to be inspected on a 6.35 mm grid using a scanner with a maximum linear speed of 500 mm/second. Ignoring the time required to accelerate and decelerate the scanner, the time required to inspect the plate is approximately 175 to 200 minutes. This time can be decreased to approximately 35 to 45 minutes by using an array of, for example, five identical transducers with multiplexed or parallel data acquisition.

The inspection time can be decreased further by using rectangularly shaped "paintbrush" transducers, which are commercially available in lengths up to 50 mm. While their large beam footprint makes paintbrush transducers relatively insensitive to small anomalies, they are commonly used for initial qualification scans because they permit much faster scanning of large areas. If for example an array of five paintbrush transducers is used and the scan index increased accordingly, then the total scan time for the hypothetical plate can be reduced to approximately 12 minutes. Paintbrush transducers have the potential to significantly reduce inspection time, but they can also introduce some challenges. In particular, they are very sensitive to alignment. If the transducer face is not parallel to the surface of the plate, the surface echo becomes distorted and can mask echoes from flaws near the surface. If the water path between the transducer and the plate varies too much, the beam intensity at given depths in the aluminum can vary and affect the interpretation of results. Techniques such as automatic water path adjustment, automatic transducer beam normalization, noise reduction algorithms, and automatic gate tracking can all be used to optimize performance when using paintbrush transducers.

After the detection scan is completed, the computer can then identify the specific regions that should be characterized. The computer can graphically indicate these regions to the operator for confirmation prior to initiating the characterization scan.

3. Flaw Characterization

The method of characterizing a flaw depends on whether the beam diameter of the ultrasonic transducer is smaller or larger than the diameter of the flaw. If the ultrasonic beam diameter is smaller than the flaw size, the beam is used to characterize the flaw by scanning the flawed area at a resolution that is finer than the beam size. If the beam diameter is larger than the flaw, it does not make sense to attempt to image the flaw in this way. Characteristics of the reflected waveform must be used to classify the flaw, which may not be possible. It is generally preferable to image the flaw directly, which means that the transducer beam diameter must be smaller than the smallest flaws to be characterized. Common transducer focussing techniques permit beam diameters of a small fraction of a millimeter. In the remainder of this paper, it is assumed that flaws will be characterized by imaging rather than waveform classification.

The time required to perform a flaw characterization scan depends upon the resolution.. For example, if the flaw was detected using a paintbrush transducer with a 50 mm width, then the area to be scanned for characterization would be approximately 50 mm by 50 mm. Since such a small area is being scanned, the scanner cannot accelerate to its full speed. Assuming that an average scan speed of 50 mm per second can be achieved, a 50 mm x 50 mm region can be scanned with a resolution of 6.25 mm in about eight seconds. If 4 bytes of information are stored for each data point, a total of 256 bytes of storage will be required. The following table summarizes scan times and storage requirements for a 50 mm x 50 mm region as a function of resolution:

Resolution (mm)	Scan Time (seconds)	Storage (bytes)
6.25	8	256
2.5	20	1,600
1.125	40	6,400
0.625	80	25,600
0.25	200	160,000
0.125	400	640,000
0.06	800	2,560,000
0.025	2000	16,000,000




To fully understand the implications of these figures, consider a typical plate that has, for example, twenty such regions that must be rescanned. If we assume that it takes approximately 30 seconds to move to each region and set up for a scan, then the following table summarizes the total time and storage requirements per plate as a function of resolution:

Resolution (mm)	Total Time (minutes)	Total Storage
6.25	13	5.1 Kb
2.5	17	32 Kb
1.125	23	128 Kb
0.625	37	512 Kb
0.25	77	3.2 Mb
0.125	143	12.8 Mb
0.06	277	51.2 Mb
0.025	677	320 Mb

Even for the coarsest resolution of 6.25 mm, the time required for flaw characterization is as long as the time required for flaw detection. For finer resolutions, the time and storage both increase considerably. Thus, it is always most efficient to scan at the largest resolution that is consistent with the visualization requirements of the application at hand.

4. Test Results

This section shows the results of some typical scans of artificial flaws in a 200 mm x 400 mm aluminum reference block. These tests were performed to show the effects of different transducers and scan resolution on flaw detection and characterization. The thickness of the aluminum plate used was 6.25 mm, and the artificial flaws were 1.25 mm diameter flat bottom holes drilled approximately half way through from the back side. Near the center of the block, a cluster of five holes was drilled with a separation of approximately 5 mm between holes. Two additional holes were located along the centerline approximately 57 mm away.

Comparative tests were conducted with a Panametrics Multiscan digital imaging system. The characteristics and part numbers of the transducers used for this test were as follows:

1. 6 mm x 50 mm element 10 MHZ narrowband paintbrush transducer (A344S-SU)
2. 12.5 mm element 10 MHZ broadband transducer, unfocussed (V311-SU)
3. 6.25 mm element 10 MHZ broadband transducer, focussed at 25 mm (V312-SU F=1)

These represent common transducers with relatively coarse, intermediate, and fine flaw resolution capabilities. The actual transducers used in a given application will vary depending on plate thickness and the minimum size anomaly that must be detected, and they should always be selected on the basis of the application's specific flaw resolution requirements.


Click for each on Image Map, or all in one Figure 1

Figure 1 depicts scan results obtained by using the paintbrush transducer at a scan increment of 3 mm (and a scan index of 25 mm, or half the element width). In these and all other images, red areas represent maximum relative signal amplitude, and blue areas represent minimum relative signal amplitude, with yellow and green in between. The second frame from the left represents a C-scan image made by gating between the front and back surfaces of the test block. In the third frame this is presented as an isoplot. The solid red imagein the fourth frame represents the results of gating the back surface echo. Because these holes are much smaller than the area of the beam, there is no significant loss of backwall signal in this case.

This image took less than one minute to make, but the tradeoff against resolution is clear. Some sort of anomaly can clearly be seen in the center of the block on the flaw gate scans, but it does not appear on the backwall scan. There is no unambiguous indication of the two small flaws on the left and right sides of the block. This type of setup would be recommended for initial, fast, and relatively low-resolution scanning of large plates.


Click for each on Image Map, or all in one Figure 2

Figure 2 shows various results using the 12.5 mm diameter unfocussed transducer at scan increments of 1.25 mm, 3 mm, and 6 mm, which required times of approximately 7, 4, and 2 minutes respectively. The decrease in resolution as increment increases is apparent, but even at 6 mm the flaw gate scan image clearly shows that there are three, not one, anomalous areas in this block. On the other hand, there is still no visible effect on the backwall scan.


Click for each on Image Map, or all in one Figure 3

Figure 3 shows high-resolution images generated with the focussed 10 MHZ transducer, whose -6dB focal spot side is approximately 0.625 mm, or half the size of the drilled holes. The first frame at left represents the image plot from the flaw gate, clearly showing all seven holes in the aluminum. The secondframe shows that the backwall gate is providing information, as all seven holes are clearly visible, as well as some machining scratches on the back surface of the block. The remaining frames represent standard and isoplotted zoomed images of the central area of the block, showing five clustered holes.

This type of scan is recommended for detailed flaw characterization, after anomalous areas have been identified by other means. Resolution is very high, but imaging is relatively slow--approximately 45 minutes to scan the entire 200 x 400 mm block at a 0.2 mm increment.

This image specifically attempted to maximize resolution. It could be run at least three times faster without significant loss of detail, especially when smoothing algorithms or other graphics processing used.


Click for each on Image Map, or all in one Figure 4

Figure 4 shows two other methods of presenting scan information. A B-scan is a cross- sectional representation of a part, while an A-scan is an RF or rectified waveform display. In figure 4, the first frame at left shows a zoomed portion of the scan made with the focussed transducer, with a section marked by a line. The second frame is the B-scan of the marked section, and the third frame shows the waveform that was stored in system memory at this point. The thick vertical colored line at the left side of the B-scan frame represents the surface echo, while the two small lines in the middle represent the position and relative size of the artificial flaws, and the thin vertical colored line at the right of the frame represents the backwall. Finally, the last frame is the full RF waveform, showing the gates used to make these images.

5. Factors Affecting System Design

There are many factors, in addition to the validity of the actual test, that influence the specification and design of any automated ultrasonic inspection system. Some of these factors are briefly described below:

• Customer Requirements: The user may impose specific ultrasonic inspection requirements based on intended use.
  
  
• Production Rates: If the ultrasonic inspection is in the production line, the speed of the inspection is critical because it directly impacts the production schedule.
  
  
• Cost of Inspection: The cost of the inspection directly adds to overall production cost. This additional cost must be justified by the benefits of the inspection, such as known product quality.
  
  
• Quality Improvements: Inspection results can be used as feedback to production in order to improve product quality.
  
  
• Inspection Results Database: Inspection results can be archived on storage media to provide a permanent record that can be correlated with any problems with the product after it is sold.
  
  
• Product Classification: Inspection results can be used to properly classify the grade of product. Optimum classification may permit previously unusable product to be sold as a lower grade, and previously undervalued product to be sold as a higher grade.
  
  
• Control of Cutting: Inspection results can be used to cut the product into pieces of different grades so that the areas without flaws can be sold as a higher grade than the areas of it that contain flaws.
  
  

Many of these benefits may not be fully realized unless the ultrasonic inspection system is carefully specified. The purchaser generally must work closely with the system supplier to balance the cost of the system with the features required to meet the inspection objective.

6. Summary and Conclusions

Ultrasonic inspection of aluminum plate products is a valuable tool for controlling and establishing quality. Using today's technology in computer based inspection systems, a higher degree of automation can be achieved than has been previously possible. For example, inspection results can be stored in a data base for comprehensive analysis, and real-time sorting and cutting decisions that enhance the value of the product can be made. It is also possible, using networked computers, to begin analyzing one portion of a plate image while the remainder of the plate is still being scanned.

For an inspection system to be cost effective, it must add real value to the product by quantitatively assessing product quality. The system purchaser must be aware of the capabilities and limitations of automated ultrasonic inspection equipment before attempting to specify a system, and generally must work closely with the system supplier to make sure that the requested specifications are achievable at reasonable cost.

Authors:

Dr. Thomas E. Michaels, Tom Nelligan, and Tim Hynes
(Panametrics, Inc. NDT Division)
Phone: +1-617-899-2719
E-mail: nelligant@panametrics.com
Homepage www.panametrics-ndt.com

The Paper was presented on the UTonline Application Workshop in May '97