A new imaging process can identify extremely small surface or sub-surface cracks in silicon nitride bearing balls.

Silicon nitride (Si₃N₄) bearing balls have several qualities that give them a performance edge over steel balls. A silicon nitride ball weighs only 40% as much as a steel ball; it is therefore subject to lower centrifugal forces and can withstand more rapid acceleration and deceleration. In addition, it experiences less vibration because its coefficient of friction is only about 20% of the coefficient of friction of a steel ball.

Silicon nitride’s coefficient of thermal expansion is also around 30% less than that of steel. When a bearing becomes very hot, clearance between surfaces is better. A Si₃N₄ball withstands extreme conditions better than steel, in part because the very low level of adhesion between silicon nitride and steel gives silicon nitride the advantage in extreme conditions.

However, Si₃N₄balls are not without weak points. In steel balls, static vibration can cause spalling and premature failure. Silicon nitride balls do not experience static vibration, but they are still subject to spalling and flaking that can cause premature failure. The cause is often a microscopic surface or near-surface crack that grows as a result of repeated stresses. These cracks generally originate at the surface and propagate at shallow depths until a spall or flake is released.

In applications where a long and predictable fatigue life is critical for a Si₃N₄bearing ball, balls may be inspected using acoustic microscope systems* to exclude those balls that have near-surface cracks. The acoustic microscope is non-destructive in its operation and can find very thin cracks that may be difficult or impossible to find by other methods.

  * Method developed by Sonoscan for its C-SAM^® acoustic microscope systems.

The Imaging Process

An acoustic microscope uses a transducer that pulses ultrasound into the sample and receives the return echoes. The echoes are reflected only by the material interfaces in the sample.

In an essentially flat sample, a material interface might be a bond between a polymer and a metal. But a material interface can also be the boundary between a solid and the air or vacuum in a gap. This second type of material interface reflects nearly 100% of the ultrasound, while solid-to-solid interfaces typically reflect much less. Acoustic microscopes find and image both types of interfaces, but the solid-to-air interface appears brighter in the acoustic image because of the higher amplitude of the echo signals.

In a flat sample, the transducer raster-scans the surface, pulsing and receiving thousands of times per second as it scans. In one spot, the pulse might strike a polycrystalline diamond/tungsten carbide interface, and 32% of the ultrasound would be reflected. This percentage can be calculated from the density and acoustic velocity of the two materials. In another spot, the pulse could strike an air-filled void, and the polycrystalline diamond/air interface would then send back virtually 100% of the ultrasound.

Silicon nitride bearing balls are handled differently during acoustic imaging by a modified C-SAM system because the balls are neither flat nor layered. Nevertheless, the basic principle—that gaps reflect essentially all of the ultrasound—still operates.

In the case of bearing balls, the transducer remains stationary while the ball is moved and rotated in order to inspect its entire surface. Each of the several thousand pulses per second is aimed at the ball surface at an angle. If there is no crack or gap at that precise location, the ultrasound will enter the Si₃N₄, travel along at and just below the surface for a short distance, and be detected by the transducer a tiny fraction of a second later. On the other hand, if the ultrasound encounters a gap or crack, it will be almost entirely reflected back away from the transducer, and no signal will be received (see Figure 1).

The pixel in the acoustic image representing this location will be black, rather than the pale gray that would be expected if the ultrasound had traveled without incident through homogeneous, defect-free silicon nitride. Surface-following waves are known as Rayleigh waves, and the acoustic imaging process is sometimes called Rayleigh wave imaging. (On a far larger scale, Rayleigh waves generated by earthquakes can travel for thousands of miles.)

Figure 2 shows the position of the transducer (the cylindrical item on the right) and the fixture holding the bearing ball (left). The northern and southern hemispheres of the ball are imaged separately. Scanning begins at the equator, with the bearing ball positioned in such a way that the pulse coming from the transducer strikes exactly at the equator. The bearing ball described here is 0.5 in. in diameter and was imaged using ultrasound at the very high frequency of 230 MHz. Higher frequencies give higher resolution; the spot size of the focused ultrasound at this frequency is about 10 microns.

After the fixture has rotated the ball 360° to cover its full circumference at the equator, the position of the ball is changed slightly to scan its circumference just above the equator. To image the whole region from the equator to the north pole, 2,048 rotational scans were made, and the ball was repositioned after each scan. Image resolution depends, in part, on the number of scans; a smaller number of scans could have been used if a lower resolution had been acceptable. After scanning half of the ball, the ball was remounted with the south pole outward, and the process was repeated.

Analyzing Results

The imaging process is primarily searching for tiny surface or near-surface cracks that can grow until they cause the spalling or flaking that can lead to premature failure of the bearing ball. If a bearing ball has no such cracks, the ultrasonic pulse will simply be mildly attenuated by its brief subsurface excursion at each of the millions of locations. The acoustic image of the entire ball will simply be a featureless pale gray. If the pulse does encounter a crack, the crack will be black because no echo was received.

Displaying the acoustic image of the whole ball, or of one hemisphere, is much like creating a flat rectilinear map of the earth: as you move north from the equator, the east-west distance becomes smaller, and is therefore stretched in order to create the map. When the transducer scans the ball’s equator, it is scanning the ball’s maximum circumference. The next scan to the north (out of 2,048 scans) will be stretched slightly to create a rectangular map. The image of a crack is therefore usually distorted to some extent. For purposes of screening bearing balls for critical applications, this distortion is unimportant; the mere presence of the crack disqualifies the ball.

Figure 3 shows the acoustic radiation pattern from a crack in the Si₃N₄ ball from Figure 2. This is a “C-crack,” one of the most frequent crack types. Note that the area surrounding the crack is uniformly pale gray, as was the image of the entire surface of the ball except for this crack.

It is not hard to imagine how repeated stresses of various types could cause the crack to expand until a spall or flake breaks loose. The stresses are so high that the loose fragment immediately causes the whole ball to shatter. The resulting fragments quickly cause the adjacent balls to shatter, and the chain reaction continues until the whole mechanism has failed. In critical applications such as jet engines, where Si₃N₄ balls tend to be used, the ultimate result of an undetected surface or near-surface crack can be catastrophic.

Because high-frequency ultrasonic pulses from the transducer cover the entire surface of a Si₃N₄ bearing ball, this method is capable of finding and imaging even extremely small surface and near-surface cracks. By removing bearing balls with cracks from assembly, the method greatly reduces the possibility of failures in service.

For additional information, contact Sonoscan at 2149 E. Pratt Blvd., Elk Grove Village, IL 60007; (847) 437-6400; email info@sonoscan.com; or visit www.sonoscan.com.