March 03, 2021 · 3 min read
In the first stages of its development, AiSight found itself at a very common crossroad. Should we or should we not build our own hardware? The easiest path is to only be a software company. In this article, we will explain why it is so and what convinced us to build our own hardware anyway.
Why hardware is hard
The first and foremost reason why hardware is hard is that the product is physical, and therefore, it can only be released when it has proved to be flawless. Software, being entirely digital, leaves a lot more room for improvement and change. Bugs can be fixed at all times, and releasing new updates is far easier than creating an entirely new line or a new version of it. This makes receiving feedback from customers much easier, and in turn, improvement never stops.
The costs of such processes are very different, too. Putting a faulty product in the market can have much worse consequences than releasing a bugged software, not only in terms of money but also of customers loyalty and satisfaction and sometimes safety.
The second reason is that hardware is much more expensive, both in terms of money and time. The iterations take longer, some of them can be costly, and it’s more difficult to predict the outcome. The first example that comes to mind is that of certification. To certificate a hardware project, it is necessary to have it tested. Testing it takes a long time, and its success depends on many different factors. In other words, it’s not enough that the product simply “works” in order to pass a test. This makes predicting the outcome of this iteration very difficult. Whenever any of these steps are not successful, it will be necessary to go back and start again.
All the above-mentioned characteristics of hardware make it especially harder for startups. Funding is a fundamental component of a startup’s daily life, and of course, investors are aware of the difficulties that hardware companies have to face. The typical funding rounds of startups are often too short compared to the iterations that products have to go through, and the ease of distributing digital product makes software startups much more desirable.
The advantages of proprietary devices
After this premise, you might wonder why would any company want to build its own hardware. We will give you some hints. Think of typical examples of companies that created their own proprietary devices. The first example that comes to mind is Apple, with its iPod and iPhone. These two products were definitely neither the first music players nor the first smartphones to appear on the market. What made them stand out, however, is how both products were able to provide an unprecedented software experience. In the case of the iPod, from when music was purchased and managed on iTunes, to when it was transferred on the iPod.
The same thing can happen with machine monitoring. Integrating hardware with software is the only way in which the data collection and analysis are perfectly optimized and translated into results that are at the same time incredibly accurate, detailed, and easy to understand.
Vibration analysis: the requirements for the perfect hardware device
All in all, what are the requirements you have to consider to make sure that your maintenance solution is the best possible? When we at AiSight found ourselves at the crossroad between Software and Hardware company, we considered all the possible alternatives to building our own device (spoiler: we’re so happy we took the difficult path, look at our Aion device now). Here they are:
- Use the internal sensors of industrial machines. Some industrial machines already have their own sensors embedded in them, and it could have been possible to exploit these to extract the necessary data to monitor performance and make predictions. However, not all machines have these sensors, and not all of them are vibration sensors (and here’s why vibration analysis is the best choice for industrial machines). Finally, even internal vibration sensors may differ across different machines, which makes adapting the algorithms much more difficult. This implies that the models have to be changed accordingly, slowing down the IT setup tremendously. Even worse, when a mix of different sensors is needed, to diversify the analysis and achieve results that go beyond simple predictive maintenance, it will be even more difficult to align all these requirements.
- Buy sensor kits on the market. At this point, it may seem like a good option to simply buy sensor kits on the market. However, this doesn’t really solve the problem. It is not easy to find exactly what is needed in a sensor kit. Often the sensors do not provide the necessary qualitative requirements, and flexibility would be incredibly limited. In the case of Aion, producing our own hardware allowed us to reach the best possible ratio between vibration and transmission, optimizing the reading window to our algorithm. Moreover, it also allowed us to achieve edge computing and run our ML algorithm on the device. Market sensors are not powerful enough, and there is not the flexibility to adapt their firmware to the highest level.
A condition monitoring solution tailored to your needs
Finally, to achieve our mission and reach unlimited machine uptime for industrial machines, we realized it was necessary to create a device that could allow our software to work at its full potential. If you want to get started with Aion, schedule a demo now.
© 2021 AiSight GmbH. All rights reserved
Phone: +49 (0) 30 40363399