Democratize AI-Enabled Machine Vision with Inspekto

Inspekto – Ready-to-Use AI-based visual inspection ·

Show notes

Tired of complex vision systems? In our todau's episode, Siggi Duell from Siemens reveals how Inspekto is revolutionizing quality control using Artificial Intelligence (AI). Imagine setting up advanced visual inspections without needing coding skills or AI experts. We'll discuss Siemens' new AI-powered machine vision system, designed for any engineer on the shop floor. Learn how it simplifies setup, eliminates project headaches, and delivers top-tier quality checks. Plus, we'll showcase real-world results from the automotive industry, proving that powerful AI is now accessible to everyone.

Key Takeaways:

Simple AI-powered vision system setup.

No coding or AI expertise needed.

Improved quality control on the shop floor.

Real-world automotive industry examples.

Making Industrial AI accessible to all.

Speaker:

Siggi Duell

Stream Lead for Industrial Vision Applications at Siemens

For more information and/or to contact our Siemens representative, click here

Transcript

Ziggy, do Thank you so much for having me today. I will show you in the next 20 minutes how everybody can do AI in the space of manufacturing without any prior knowledge in the space of AI and unlocking great value in production with this. Let's dive right in. What is Inspector? I'll be presenting Inspector today, our new system that allows anybody without prior knowledge in AI and machine vision to do just that. Machine vision to improve quality in manufacturing in many, many different areas without prior knowledge in the field of asset AI and also machine vision. In the next minutes you will learn how one can do that, and I'll show you actually how and do a demo on stage in the next video. Before we come to inspector, however, why would we do inspector and what is so different and unique about an inspector? When it comes to machine vision, a lot of applications are typically around very complex projects. They require know-how. Typically the timeline is very difficult to estimate. A lot of expertise is required, and putting all of this together into a working solution is typically a one-off that can be potentially rolled out on one line, on one machine, on one type of application, but it's in most cases use case specific. And that's the starting point for most machine vision applications today. When it comes to advanced machine vision, however, we want to You want to tackle this by bringing in Inspector. With Inspector, we are aiming to reduce the effort to do this dramatically. So we see a lot of projects that recently have done in a matter of weeks, we can bring this down into days. When it comes to predictability of the total cost of ownership, we see a dramatic increase in viability for our customers with this. And as I said before, the most important topic is we don't need expertise in the space of AI and machine vision. So let's show you how that works. The first thing that we need to do when we do machine vision is we need to learn how to set up the process, how the part that we inspect is actually looking like. Here in this example, we have a PCB, a printed circuit board, and in this case one wants to inspect in this case, certain parts on that board, certain chips if they are there. So we want to make sure that they are present. We want to make sure that any sort of components are oriented in the right way, so they are actually not oriented in the wrong way. In this case, Capa just as an example. You want to make sure that there is no soldiering defects on this part, so we want to make sure that all of this is actually in the right place and typically our customers and also we, when we do this in our own factories, we know how our products look like, so we can do this and actually anybody qualified in the process can do this without prior knowledge in the domain. The next steps after we told the system what our part is, in this case this printed circuit board, is setting up illumination and setting up the lighting on this. And for anybody that did a project in this space before, that's typically a work of art. It's very, very difficult to get a pic picture, get an image that is not full of reflections that actually has the details that we are focused on exposed the best and in many, many. Cases one which send parts to a certain solution provider they would take this into the lab they will try different lights and different illuminations, different filters, different lenses, and different cameras with this to end up with the best possible image for you. Just saying this takes quite some time, right? With Inspector, all you need to do is actually press one button. So that's the button, and after you press that button to optimize your image, the system itself will test many different combinations of lighting. It will test many different directions of lighting because we have a multidirectional light within Spector. It will also test different camera settings, all of this combined to get the best possible image here. Inspector is actually doing a lot of computational photography here. If you want to see this later, you can also see this uh actually yourself in action. So once you did this, you have a good image. And with this image you can now go back and have done what typically takes a project, so have a good image and use now again inspector in the same UI to now set up your use case. And how you do that is very simple. You take 20 samples, 20 good samples that you place under the machine, and you actually take them. With those samples in, you pretty much have finished learning the use case. So that's very, very straightforward. Anybody can do that. If it's a manual process, you simply put them there. If it's an automated process, you simply start your conveyor. You can also do this in motion. You can do this with a robotics application. So that's done in a way that this would not require again any project, but anybody in the space of factory automation or for manual stations, the workers can do that kind of job. And after you have done that, the next step is simply use the system. Again, it's the same UI, and this takes a matter of minutes to set this up. So in the sphere I took now 5 minutes to present to you what in reality might take actually 10 minutes, but it's actually not that much longer to get to that point. And this is not limited to BCBs, so this can work for plastic parts as you see, same way. You can inspect the defects also in the same tool, so you don't need to leave the program. You can immediately see if something is defective. If you have metal parts again they can be very reflective, you can also inspect that with Inspector. You have all the same abilities to track and trace your parts again all in one software. And with all of those knowledge in hand, this opens a very wide space of use cases with very, very simple tooling and a very, very eas easy application that you get actually out of the box. When we look at our customers, what they are doing with this is many, many use cases in the space of electronics, in the space of discrete manufacturing of metal parts, plastic parts, assemblies. There can be also complex packaging processes where we want to make sure that parts are optimal in condition, they don't have any surface defects, that if you have an assembly that everything is there, that all your nuts and bolts are in, that you don't have any delicate surfaces that are defective, that you don't have any ceilings missing, that you don't have any issues with seals, if you have electronic parts, that there are no parts missing. Again, this can be done in multiple stages of manufacturing all the way to assembly and all with the same system, all out of one box and very, very important, all of the things that I've shown you right now is done locally. So when you would start with Inspector, you get all of this functionality within one box with no connectivity required, but of course, and that's very important, you have a lot of connectivity in Inspector. You can start out putting this on a bench and start inspecting right away, but if you want to connect this into our systems, Inspector brings a lot of connectivity into the Siemens universe. Not only that, inspectors open. If you come from the opposite end and say my shop floor is full of automation with other systems, that's not a problem either. We have Ethernet IP integrated at the same time you can export the data using standard IT interfaces and bring this all the way up to your cloud. If you want to document your images, you can take them off the system. If you don't do this, you can also back up the entire inspector system and simply use the inspector system to work with your data, so there's no need for a secondary system, even if you want to do track and trace. And with all of this, we get a very interoperable system that is extremely flexible for many, many different applications, very easy to use, aiming to allow to use AI for the shop floor worker, for the quality engineer without the need to need um expertise or a dedicated project to do AI here. So when we put all of this together, what we end up with is a system that is made for manufacturing that is required intrinsically to be flexible and that is ever changing. And we see this actually a lot also in our own manufacturing. With inspector, if a problem comes up in quality, there is no need for a complex project. One can simply take up an inspector, put it up there, and within half a day has not only set up the system, but also the integration of the system into robotics applications or a line if that's already available there. Since this is quite fast, there's a dramatic reduce a reduction in effort right now for this. So in contrast to a project setup that is itself expensive and just tuning a light as an example for a project may take weeks, the entire system is actually much cheaper in contrast to a project. And with its quick setup again, if there's a change in the demands that you have, you can actually very quickly not only set it up, but also adopt it to new parts. So if you have a change in your manufacturing, if you make a PCB in a different version, if you actually have a different part that just looks slightly different. Everything can be adopted by the workers on the shop floor. There's no need to send samples out. There's no need for another project or an iteration on the system. Fine tuning the system can be done out of the box by the people that work with the system, so there's no need for anything else than just the standard effort and the standard skill that you actually have in setting up your lines, setting up your new batch of product, setting off your one-off production that you might do, and you can still do quality inspection with the inspector on this. When we put all of this together, how does the Inspector look in action? We see Inspector being used on cobots as an example. It can be very easily robot mounted. Putting it on a bench with a robot, putting on a conveyor, inspecting parts at a speed of up to 0.75 m per second is no problem for an inspector. We see multi-camera setups in the field. We see integration with HEVs in complex production cells doing many inspections at the same time. Again, since the system is so easy, integrating multiple different use cases in one go also can be achieved. Even if you have started with a single use case and you see 10 other problems next to it, with Inspector, it's all something that you can program yourself. There's no need to get another project. If you see. That you started to inspect bolts and you also see there might be surface defects, you can very quickly adopt the software to do that as well all yourself. There's no need to know this upfront since it's software defined and since you can actually do this yourself without any programming, that's very easy for you to actually do yourself. Now we put all of this together and ask the question again, now what is Inspector? I promised at the beginning it's a very flexible inspection system. It's powered by AI and it's actually not only using AI for the inspection part, it's using AI pretty much for every step along the journey. In contrast to other systems that we actually also have, this is the easiest way to start with this, because anybody can in this case use AI without being an expert, and what this means is that you actually don't even see AI or you don't even feel if you go to our booth and actually look at that, you don't even see anything about AI. It's just helping you along the way. It helps you to set up the light. It helps you to set up the camera. It actually helps you to set up the use cave itself all locally on the machine. And if you actually have any sort of improvement topics, for instance, if your quality is not good enough, your part is changing, the system is actually also recommending to add more data to it all by itself. It's still very data efficient, so starting with 20 samples, but most use cases we see, the system might actually recommend you another 5 or 10 samples to reach optimal quality. But there's no need for you to figure out which samples this should be. The system is actually telling you what data to add. So with this we add actually. A very holistic system that comes with a Siemens IPC that comes with a camera and a light in the box today and all the cables, so it's ready to set up out of the box today. And gives you a very, very powerful solution to actually start in automation. Just with automation know-how and Visual domain know-how, you're able to do a visual inspection and you don't require additional know-how. That's our biggest point for Inspector. We have other portfolios that you'll see in a second, but that's really why we believe Inspector is such a great tool for us and actually also for our customers like you are. If you now ask, OK, what is my return on invest for something like this? For those who have done projects in the space of visual quality inspection, you see a lot of difficulties typically estimating this. With Inspector, since it's so easy to set up, this is actually one of the most important points. You see, it's very straightforward because all the value comes in one box, and what you have in addition to this is your typically your own cost of integrating this to a line and actually setting this up to, for instance, a robot and setting up the use case that typically takes less than a day to actually get started. And that's something that everybody can estimate quite easily since this is also done with own staff and no extra effort usually that you otherwise would have externally. Again, very straightforward to estimate your return on investment for such a solution. Also, and this is what I would like to close on this since it's so straightforward, it's very imminent. So if you start with inspector today, since there is no project, you can actually get this value today. And this is not only for the first application we see also see in our own factories, if there's any changeover on inspector system, since it's so easy to change, it's very easy to also take the system and take it to the next batch, take it to a new kind of product, take it to a new product variant, or to a new problem setting altogether. Again, if you build a typical project, and that's what we typically start from, that's built for one application and one application only. If that application ends, you also need to literally come and do this assembly project, or actually refit a new project to a new problem, a new line or a new product that you make. So with this, I would like to actually invite you to see this in action. We have Inspector right behind us. You can actually see this live. You can actually see what I've shown you on videos at the booth. It takes like 1015 minutes to see an ent and demonstration on Inspector on some new parts that they have around, uh, on metal parts, but also on our own components where we can demonstrate Inspector. And with this, I would like to close and thank you so much uh for being here today.

Full transcript (WebVTT)