Design of Ultrasonic Flaw Detection System Based on PC-based Modern Operating System and Signal Acquisition Card

“Manufacturers of ultrasonic flaw detection systems have traditionally used analog circuits for non-destructive testing of thin steel components. But what the market needs now is a modern PC-based operating system that can provide higher performance and lower prices, as well as high-speed and high-resolution signal acquisition cards.

“

Manufacturers of ultrasonic flaw detection systems have traditionally used analog circuits for non-destructive testing of thin steel components. But what the market needs now is a modern PC-based operating system that can provide higher performance and lower prices, as well as high-speed and high-resolution signal acquisition cards.

Before discussing the role that signal acquisition cards play in flaw detection, let’s review the entire system. The broadband ultrasonic sensor is automatically positioned in front of the part under test, and both the sensor and the part are placed in the sink. A sensor operating in reflection mode emits and detects ultrasonic waves. After receiving excitation from the ultrasonic pulser/receiver, the sensor ignores part of the bandwidth of ultrasonic energy pulses and receives echoes reflected from the part under test.

During the extraction of continuous ultrasound recordings, the transducer was moved in 0.1 mm steps by an automatic positioning arm controlled by a GPIB bypass link. The arm scans through a 500mm x 500mm rectangular grid with a raster-like pattern front and rear, and has a fast axis and an orthogonal slow axis. While the sensor is translating along the fast axis, the positioning system trails the conveyor at a pace of approximately 1 step per millisecond.

The positioning controller is equipped with a pulse generator that outputs the position. Once the target position is determined, the controller generates a TTL pulse. This pulse is often used to cause the ultrasonic pulser/receiver to deliver an excitation. In this way, the ultrasound examination is performed automatically only at the target location.

After excitation of the sensor, the corresponding sequence of ultrasonic echoes needs to be returned because the ultrasonic waves have to travel through a waterway of about one meter long. After ultrasonic excitation, a programmable delay was used to create a TTL excitation lasting approximately 700 μs. This pulse triggers a high-speed high-resolution signal acquisition card. In contrast, the ultrasonic echo sequence lasts about 100 μs.

The role of the signal acquisition card

An embedded high-speed, high-resolution signal acquisition card is used to capture ultrasonic signals from the sensor electronics. The system combines two kinds of signals: the signal from the sensor center frequency up to 10MHz, and the sampled ultrasonic signal at 10MSPS. This sampling rate is up to 10 points per signal period, and the echo time allows excellent resolution. In order to detect the smallest possible slit echo, the largest possible dynamic range and vertical resolution are required.

The Gage CompuScope 14100 signal acquisition card provides 100MHz sampling rate and 50MHz analog input bandwidth to meet the needs of ultrasonic sensors. Figure 1 is a schematic diagram of the system.

The ±1V signal output from the ultrasonic pulser/receiver is directly connected to the BNC input of the digital card. The input impedance of the acquisition card is 50Ω and programmable, providing an input terminal that is compatible with the 50Ω impedance of the BNC line, and eliminating the distortion caused by multiple signal reflections. The output of the delayer is connected to the BNC external trigger input of the capture card. The acquisition card provides 14-bit sampling accuracy. Since the detected fracture echoes are arbitrarily small, high sampling accuracy is necessary in ultrasonic nondestructive detection.

Figure 2 shows a real ultrasonic signal from a laminated steel part. The image depicts a larger echo reflected from the front wall of the part, followed by a smaller echo, indicating that behind the surface is the crevice. The time difference between the echo of the fissure and the echo of the front wall is related to the depth of the fissure as follows.

D=vΔt/2

Among them, D is the depth of the crack, Δt is the time delay of the front wall echo, and v is the wave speed of the ultrasonic wave in the steel.

The purpose of an ultrasound scan is to determine Δt over the entire scan and to create a map indicating the depth of the cracks across the part.

The amplitude of the tracking echo increases with the size of the fracture. The amplitude of the overall ultrasonic signal is adjusted by the pulse generator/receiver gain. In this way, the front wall echo almost fills the input range of the signal acquisition card, which is ±1V in this example.

As a result, on the premise that the front wall echo does not have clipping distortion, the slit echo cannot be further amplified. Figure 3 shows the magnified image of the slit echo in Figure 2. The upper waveform is at 8-bit resolution and the lower one is at 14-bit resolution.

The amplitude of the slit echo is only about 1% of the amplitude of the large anterior wall echo. The 8-bit signal acquisition card divides the input range into 2^8 or 256 different levels. This shows that the echoes of the ladder in Figure 3 differ only by 2 to 3 steps. The echo waveform is severely distorted, and furthermore, if the echo were small, it would not be detected at all.

In contrast, a 14-bit signal acquisition card divides the input range into 2^14 or 16 384 different levels, and the slit echo spans 150 levels. As shown in Figure 3, the high resolution of the 14-bit capture card reproduces the shape and location of the slit echo. Even if the echo and background noise can be compared, the time delay Δt can still be extracted by digital cross-correlation analysis. Clearly, high resolution is crucial in the detection of crack echoes.

In a linear scan along the fast axis, the ultrasound triggers are distributed at a regular 1 kHz rate. The capture card should not miss any of these triggers, otherwise there will be no correlation between the captured waveform and the sensor position.

The time required for a linear scan along the fast axis is (500mm/0.1mm)/1kHz, or 5s. The start of the next fast axis scan is programmed; however, for technical reasons, the repositioning of the slow axis takes at least 0.5s.

The 14-bit signal acquisition card can transmit data at a stable rate of 100MB/s through the PCI bus. Therefore, the high-speed acquisition card can capture the 50μs ultrasonic wave and send it to the memory through the PCI bus for the next 1kHz trigger. The capture card meets this performance benchmark only under single-task operating systems such as MS-DOS.

Acquisition depth

There is one more problem in the Windows environment. Multitasking Windows is not a real-time operating system, so in the process of processing a task, the amount of time taken up by Windows services other tasks is uncertain, so there is not enough waveform capture capability in the Windows environment. In the fast sweep of the system, reliable waveform capture capability is very important, and no trigger can be missed.

The solution to this problem is to use an ultra-deep onboard capture buffer. Operating the data converter in multiple recording mode, successfully extracted waveform data is stacked in the onboard acquisition buffer. Between extractions, the data converters are reformatted by hardware without CPU intervention. So, once started, multi-recording mode does not conflict with the Windows environment of multi-tasking mode.

The signal acquisition card will require sufficient onboard acquisition depth to hold data for the entire fast axis scan. To determine the amount of memory required, it is necessary to calculate the number of samples required for a single-channel 100 μs ultrasound recording.

Record length=100μs×100MSPS=10 000s

Since the position step length is 0.1mm and the length of the fast axis is 500mm, there are a total of 5000 step positions in the linear fast axis scan. At each step, the signal acquisition card must capture a 10 000s record.As a result, the onboard acquisition depth requires at least

5 000 records×10 000s/record = 50 000 000s

That is, a high-speed high-resolution signal acquisition card with an onboard acquisition depth of up to 1GS is required to meet the needs.

At successive fast axis scan intervals, the system will load data from the last fast axis scan into PC memory. Capture cards can use a technique called PCI bus control to transfer data over the PCI bus. With this method, no CPU intervention is required during data transfer. In addition, the sustained transfer rate of the capture card can be as high as 100MB/s.Because each 14-bit sample occupies 2 bytes, the full data transfer for a fast axis scan requires at least

2B/s×50 000 000s/(100MB/s)=1s

In this way, the data transfer will not seriously delay the data of the next fast axis scan, because the system has already required 0.5s for mechanical stability recovery. If the data transfer process is temporarily interrupted by Windows, the transfer time will increase slightly, but no data will be lost, and once resumed, the transfer process only needs to resume where it was interrupted.

Windows 2000 tasks are developed in the SDK under the C language environment, and this tool provides convenient and easy-to-use routines to meet new start-to-customize Windows tasks.

Because this high-speed high-resolution signal acquisition card is a plug-and-play PCI device, and the underlying hardware details are manipulated by Windows, it does not require any underlying hardware programming. The Windows program starts the position motor, controls and scans the component under test, and then calls the C language subroutine to extract and transmit data from the acquisition card.

High-speed and high-resolution signal acquisition cards based on high-performance PCs can provide high sampling speed, high vertical resolution, deep acquisition depth, and fast data transmission, making it possible to build an automated, low-cost nondestructive detection system.