P. Kannan - Soliton Automation Private Limited
M. Senthil Kumar - Soliton Automation Private Limited
J. Rajeswari - Soliton Automation Private Limited
A typical instrument cluster for motorbikes contains about 10 to 15 lamps or bulbs that do everything from illuminating the cluster to indicating which gear the driver has selected. During assembly, the following faults can occur:
The wattage of the bulbs in a cluster varies depending on the brightness required for the different indicators, and an operator may interchange the bulbs during assembly.
To obtain the required color for the indicator, the operator places a colored plastic cover over the bulbs. However, he may place the wrong color over a bulb or interchange two covers of different colors.
The wiring harness that connects the bulbs to the various switches can have loose connections or the operator may interchange the wires.
The operator may assemble fused bulbs into the cluster.
Such faults cause a bulb not to turn on when switched, cause the wrong bulb to glow, or cause it to glow with the wrong color. The combinatorial probability that such a fault will occur and go undetected during manual inspection is quite large. Therefore, PRICOL, a large manufacturer of instrument clusters in India, wanted to fully automate this inspection process.
Integrating and Automating a Two-Part Manual System
In the existing system, the inspection was split into two stations. At the first station, a PLC-based system checked the wiring harness. This system could automatically do the switching and identify if there was a fault, but could not always locate the fault. In addition, the PLC-based system was cumbersome to reprogram when a new type of cluster was introduced.
At the second station, an operator manually inspected the lamp by sequentially switching through all the connections while looking at a master cluster and the cluster under test. By comparing the lamps that turned on in the two clusters, the operator could verify that the lamp that lit up was in the right place with the right brightness and color. When doing this test repeatedly during an entire shift, there was a higher likelihood for the operator to make mistakes and the determination of the brightness was only qualitative.
Our first objective was to integrate the two tests into a single system. Both tests involved switching and interfaced to the cluster through the same connector; therefore, it was natural to combine the two tests into the same station.
Although the automated system does not need a master cluster for comparison or reference, PRICOL wanted us to integrate this in the event that the automated system had a breakdown and they needed to perform the tests manually as a backup.
We interfaced a low-cost 640 x 480 resolution color camera to an IMAQ 1411 color image acquisition card. Using the switching circuit, we switched the 12 V power to each of the lamps and took an image for analysis. Since the cluster was located in a fixture during the test and we did not have to consider any movement, we used a fixed region of interest (ROI) to determine if the correct lamp was glowing. The intensity within the ROI had to be in the specified range (determined during the setup) and we verified that the intensity of all of the other lamps was lower than a specified threshold. Simultaneously, we measured the current drawn by the lamp using a shunt resistor and an analog input channel of an NI DAQCard-6024E, and we verified and recorded the wattage.
The color of the lamp was harder to determine than we had anticipated because the camera did not have exposure control and when the lamp was switched on in the dark, the light saturated the CCD elements and blooming occurred in a small region surrounding the lamp. We noticed that the region where the blooming occurred contained the color information. Using NI Vision Builder and many images, we determined an algorithm that provided the color of each lamp using the information from the surrounding pixels because the standard color-processing VIs did not yield the correct answer.
Because the connections for lamps in the cluster (including, illumination, hi-beam, and turn) were nontrivial, we had to program in a good deal of logic to isolate the fault in the wire harness when there was one. Space does not permit us to describe the details.
Reducing Operating Cost and Allowing Quick Process Corrections
Using NI LabVIEW and the IMAQ Vision tools, we developed and validated the software in six weeks and the entire system in eight. The fully automated integrated cluster inspection system proved to be vastly superior to the older system that required two test stations and two operators. Manual inspection was prone to human errors, but our integrated automated system offers foolproof testing. In addition, our system requires half the space and only one operator to load and unload the clusters, reducing operating cost to less than half.
Unlike the PLC-based system, which was not easy to reprogram for new cluster types, our new system based on LabVIEW provides easy reprogramming. Previously, PRICOL had to conduct separate offline analysis to find and correct a fault. Because our automated system identifies the source of the fault and provides an online Pareto chart, it saves time for the rework team and allows quick process corrections. Finally, PRICOL now can verify bulb wattage through current measurement and determine brightness quantitatively from image analysis.
Soliton Automation Private Limited
Tel: +91 (422) 2302374