Creating a System to Analyze Air Bubbles in Liquid

Author(s):
Ernst Wilding - Joanneum Research

Industry:
Machines/Mechanics, Machine Vision/Imaging

Products:
Machine Vision, LabVIEW

The Challenge:
Analyzing air bubbles in liquid in an industrial setting.

The Solution:
Using a powerful LabVIEW development system to build a statistical analysis tool based on image processing.

In industrial processing, the size of air bubbles in liquid respective to their size distribution plays a major role in the effectiveness of the production process. A range of bubble-generating equipment exists, and each differs in how it produces bubbles and releases them into the liquid. Some of these differences include bubble speed, amount, and distribution. Previously, only analysis tools existed that qualified single bubbles, revealing the problem of a missing statistical relevance. Engineers could only achieve this relevance through the time-consuming and work-intensive process of measuring a large amount of single bubbles.

Our development system results in a reproducible qualification of generated bubbles anywhere in the observed liquid based on a statistical random sample test. We developed a powerful imaging sensor that you can freely place in the observed liquid, image processing software, and application user interface.

To achieve statistical qualification of bubble-generating equipment, we needed to develop a new imaging sensor that would make it possible to acquire images of the emerging bubbles anywhere within the liquid.

Because of the relatively high speed of the bubbles leaving the bubble-generating equipment, the image acquisition task was difficult. The purpose of the demanding image processing system was greatly influenced by the high density of the bubbles in the medium and the light refraction and reflection.

We created a system based on an industry-standard personal computer running Microsoft Windows NT and the imaging sensor, which includes the mounted backlight unit with adjustable light intensity.

Developing the Imaging Sensor

Initial image acquisition tests showed that depending on the type of illumination (direct light or backlight) bubbles and their shapes appear different. These tests also revealed the high density of bubbles, which makes it difficult to distinguish between individual bubbles. Partial or total overlay of bubbles presented another challenge.

With these considerations we developed the new imaging sensor. It consists of a camera mounted in waterproof steel housing oriented 90 degrees from the viewing direction. The image is redirected to the camera using a prism. This configuration provides a slim design that makes it possible to use the sensor even in a narrow experimental container.

We can freely adjust the light emitting diode (LED) backlight used in relation to the sensor window so that only a small gap remains where the bubbles can pass between the illumination unit and the imaging sensor. This helps the system work with varying bubble densities. The intensity of the backlight is also adjustable to overcome light intensity problems in fluids that are not fully transparent.

Two Methods for Image Processing

Several morphological image processing steps, such as morphological gradient, morphological filling of holes, and morphological opening, are necessary for image preprocessing and bubble segmentation. To minimize false alarm rate in the detection of bubbles, we can perform preliminary analysis of isolated bubbles in two different ways.

One way to avoid detecting multiple smaller bubbles with partial overlay as one big bubble, which would lead to wrong statistical data, lies in the use of heuristics. Heuristics increase the power of single bubble identification. A brighter center spot and a somewhat uniform but darker border region characterize the image of a single bubble. Using this information, you can reject ambiguous bubbles.

The second method is to use characteristic parameters such as the Heywood circularity factor, Waddel disk diameter, center of gravity, ellipse orientation, and a set of given parameter ranges for cross-checking the bubble shape. You can choose both methods from the user interface – their use depends on the nature of the occurring bubbles.

Functional User Interface

The user interface consists of the following functional parts:

• Image acquisition including image source and image processing parameters
• Image display source image and different stages of image processing
• Histogram mode
• Histogram display including mean, standard deviation, and bubble count as numeric output
• Mode of integration including a single image, image sequence
• Output of histogram data and statistics on serial port

We can acquire images from the sensor in individual shots or as a sequence of images. We can also load images saved during earlier sessions from hard disk to recalculate statistics. Several different modes of operation determine the statistic output of the program, which we can perform after each processed image or at the end of an image sequence. In addition, we can save histogram data on a disk. We can adjust image processing parameters to optimize the processing result for the given situation of bubble speed, liquid transparency, and bubble size.

Moreover, histogram options enable selection between different desired histograms that include the following:

• Absolute quantity frequency density of bubble diameter
• Relative quantity frequency density of bubble diameter
• Relative surface frequency density
• Cumulative frequency of quantity
• Cumulative frequency of surface

Rapidly Implementing a User Interface and Reducing Development Time

Our application qualifies individual bubble-generating equipment with respect to generated bubble amount, bubble size, and size distribution. The newly developed innovative imaging sensor provides the statistical analysis anywhere in the liquid. The system can measure bubbles with a speed up to 1.0 m/s and a spatial resolution of 0.2 mm. With the new sensor, you can measure the bubbles directly above the equipment where the bubbles are generated, and during movement toward the liquid surface.

Using NI LabVIEW, we rapidly implemented a functional and clearly structured user interface and significantly shortened the development cycle compared to similar projects.

Author Information:
Ernst Wilding
Joanneum Research
Andritz AG
Andritz Austria 64
Austria
Tel: +43-316-875-1740
ernst.wilding@joanneum.at

Browse All Case Studies »