Manufacturing ADAS Cameras Calls For Analytics

Create automatic rules based on line data to proactively alert and react to problems before they result in scrap.


Automotive camera design and manufacturing is one of the most difficult undertakings in hardware. Not only do you need to deal with the usual difficulties of high performance mechanical and electrical designs, but now highly specialized and sensitive optics are added into the equation. Complicating factors is that you won’t know how good the product is until it’s been completely assembled, while also having little way to correlate that to real world performance, and even with such a correlation rework is not allowed on the devices. All these factors go into creating expensive products with high potential scrap rates that are extremely sensitive to process and material variations. So what do you do to try to mitigate these issues, making the most cost effective and highest quality product you can? One answer lies in the power of big data analytics.

There are many different big data analytics platforms, but there are a few key areas that are often overlooked when looking into finding the right solution to completely harness your data. The first is the acquisition, preparation, and centralization of the data being collected off the machines in the line, in addition to harmonizing the data so that you have traceability between the diverse data types and sets. This is more challenging than it seems, because the data coming from each machine is very different, so it needs to be prepared so that it’s in a useable format for the data analytics, one of the largest challenges for data scientists. Beyond that, you also need to be able to gather data in real-time so that you’re not working on old data and are able to react to changes more quickly. Then once you have the data, it needs to be centralized so that it can be readily and quickly accessed and used by the different teams. The data doesn’t just need to be available while the sensors are online but also accessed while offline so that historical trends can be closely analyzed to create a better view of the data. Essentially, you want to be able to easily and quickly access and correlate data from Final Test to CMAT to incoming material inspection in real-time so you can derive rapid insights.

Once you have the data and it’s prepared for consumption, you now need to be able to run the analysis and experiments to try to get to the root cause of any issues that may be on the line, or to monitor the health of your processes. For this, you need to be able to quickly react to the data coming off the line, as working on old data will not be as effective as understanding the current state of your processes. This could be looking at the MTF of your incoming lenses to select parts that you know will make it through the build successfully, or being able to correlate field data back to the incoming inspection of the lenses and image sensors, so you have a higher confidence that the parts are built right the first time, reducing scrap. Analysis without action is not useful though, so you need to be able to create automatic rules based on the analysis that monitor the lines to proactively alert and react to problems before they result in scrap.

If a problem is found, you need to be able to find the source of the issues quickly using techniques like multivariate analysis. This allows you to be able to look at all the variables, whether it’s cure time of an epoxy, the MTF of incoming lenses, or the particulate matter in the air, you can run quick analyses to get to the root of the problem. Only by having access to all this different data and having it in an easily accessible format will you be able to get to answers on the line quickly. The longer it takes to get to the answer, the more money is lost through scrap or missed shipments, again reinforcing the need to have automatic rules that detect issues before they become built in problems.

Over the many years it’s been in business, Optimal+ has come to understand all these challenges and has created an open platform that specifically addresses the problem using big data in a factory setting. Specifically, we connect all your machines to the platform and automatically prepare, cleanse, harmonize, and store the data in real-time so you always have access to the most up to date and complete data set, enabling both online and offline big data analytics. Our powerful analytics and rules allow you to not only monitor the health of your line in real time, but to quickly drive down and connect data that you may have not been able to before, such as connecting the MTF of incoming lenses to Final Test performance. Being able to create and deploy rules that run in real-time allow you to prevent scrap by reacting early in the process. Additionally, you can reduce the amount of scrap in your line by only allowing optimal parts through the line based upon ML algorithms and rules through our platform. This enables the ability to use prescriptive and preventative analytics and rules instead of reacting products after they have arrived at end of line. Over our long history, Optimal+ has created tools to not only speed your ability to quickly access and analyze your data, but also enables you to reduce your scrap and improve inline process automatically so that you can focus on making the highest quality and performance products on the market.

Leave a Reply

(Note: This name will be displayed publicly)