Anand Chinnaswamy - Soliton Automation Private Limited
P. Kannan - Soliton Automation Private Limited
A. Ashok Kumar - Soliton Automation Private Limited
Our customer is a Tier 1 automotive supplier of brake drum assemblies with customers both in India and abroad. In the past, they have tried a number of different vision systems for brake drum assembly inspection, but have had very limited success. The main problem with the previous systems was their dependence on using pattern matching, which proved inadequate in handling the wide variations and complexities in the assemblies.
The test requirements call for the detection of:
- Presence of various components
- Reversal of springs
- Proper locking or engagement of components
- Direction of assembly of geared components
- Angular orientation of clips
- Position of brake linings
- Presence of threading in certain components
- Presence of lettering
- Gauging of diameters, lengths, and thickness
Our customer was looking for a cost-effective test system that could meet the following requirements:
- Flexibility – Ability to test various models of brake drums
- Reliability – Reliable and consistent inspection results
- Networking – Test results accessible over the local network
- Delivery – Aggressive 10-week delivery schedule for complete inspection station
- Compact footprint – Efficient utilization of valuable manufacturing floor space
We chose a 1280 x 960 pixel Sony FireWire digital camera with a number of programmable features, including more than 12 parameters, such as selection of shutter speed and filter, that we could configure from the application software. We used NI-IMAQ 1394 driver software from NI for interfacing with the camera.
From the beginning, our customer was very clear that they needed a vision system with complete flexibility. Based on previous experience, they were convinced that standard configurable vision systems were not suitable. However, a demo of the National Instruments Vision Development Module completely convinced them that this was the right product for their application.
Combining Hardware and Software for Accuracy and Reliability
The brake inspection station consists of a rotary indexing table with a fixture mounted on it. Each type of brake assembly has a corresponding fixture. After we manually mount the brake assembly on the fixture, a pneumatic clamp prevents the assembly from moving. A centralizing mechanism makes sure that the brake shoes are centralized before the imaging process starts.
We used a high-resolution IEEE 1394 camera with a corresponding PCI card. A software-controlled motorized zoom lens gives our client access to the full resolution of the camera for different brake models. We also provided polarizers to reduce the effects of glare from highly reflective components, and we chose high-frequency lighting sources whose intensity we could control through software.
The rotary table guarantees that all four quadrants of the brake assembly can be imaged separately so as to increase the available resolution. The images thus captured are transferred to the PC for analysis. The custom-built software analyzes the images using various techniques and actuates the marking mechanism based on whether the drum is accepted or not.
Software played a critical role in the success of the system. As mentioned earlier, some visual variations in components are considered normal for an “under the hood” application. In addition, the presence, absence, or shift in position of the components in the background affects detection of components, and shifting of the brake shoes, though limited to an extent by the centralizing mechanism, also poses a challenge.
We built custom algorithms because pattern matching was unreliable for the conditions mentioned above. For example, we designed one algorithm to detect reversal of a spring. Every time we created a new algorithm or altered an existing one, we had to validate it over hundreds of brake assembly images to ensure that it was reliable over the entire sample set. National Instruments Vision Builder, with its batch processing capabilities, proved an indispensable tool to accomplish this task.
Though we could not develop and verify new algorithms using this method, we realized that they were not 100 percent reliable due to the variations mentioned. To be certain that the system worked well, we used an adaptive technique where we iteratively changed certain parameters in the algorithm, such as threshold value until we found the feature we were looking for. If at the end of the iteration we did not find the feature, we concluded that it was not present. Based on the images from the sample set, we were able to identify a range that we could use for these iterations, thereby reducing the processing time required. We had to change threshold values iteratively from 50 to 100 in steps of five until a single particle of a certain area remained.
Fully Automated System Adapts to Wide Variations in Components
We built a fully automated brake inspection system using the latest in virtual instrumentation and machine vision technology during a period of 10 weeks. We used custom-built, self-adapting algorithms to ensure reliable inspection irrespective of the wide variations in the components. Our system based on NI products also gives our customer complete flexibility in choosing both which features are inspected and the inspection criteria, so it is very easy to change from one model to another.
Whereas defect analysis was not possible with other configurable machine vision systems, with our system our customer can use statistical details on various defects to curb problems in raw materials and upstream processes at the source. Finally, our system makes customized reports accessible over the network.
The power of the virtual instrumentation software is evident in this application. The programming capabilities of the NI LabVIEW development platform along with the vision development tools from National Instruments proved to be the main reason behind the success of this system.
A. Ashok Kumar
Soliton Automation Private Limited