FAQ

What exactly does ai-omatic solutions do?


ai-omatic offers a software as a service solution in the area of ​​predictive maintenance. The software is based on an innovative method developed by ai-omatic which, using a complex mathematical model, can recognize and evaluate anomalies without previously defined limit values. ai-omatic’s goal is unattended, fully automatic monitoring of production processes that really does justice to the term predictive maintenance.




What does the process from unstructured data to predictive maintenance look like?


The first step in the process is a workshop with ai-omatic, in which the specific use case is discussed and goals for the project are set. This is followed by a quality check of the existing data to determine whether it is suitable for the application. If there is no data available yet or if it doesn’t meet the quality requirements, data must first be collected. ai-omatic then uses this data to train a model. The next step is a two-month proof of concept, in which the basic feasibility of the project is demonstrated. The process then changes into a monthly license model. The status of the machines can be viewed continuously via the user-friendly dashboard, failures can be predicted, and maintenance costs can be reduced.




In which regions can the software be used?


The software is not limited to any single region or country.




Who is the software suitable for?


ai-omatic’s software is aimed at all companies that have an interest in preventing unplanned machine downtimes. The software is particularly interesting for such companies, where failures lead to major cost and productivity losses.

The primary target group are machine manufacturers and production companies that use machines whose complex systems produce large amounts of data and whose regular operation is constant over long periods of time.




Which kind of machines is the software suitable for?


ai-omatic’s software is suitable for all kinds of machines that have mechanical motors or parts that rotate. In order to be a use-case the monotoring of those sensors must produce large amounts of data. As an example, ai-omatic works with machines from the automotive, shipping and manufacturing sectors.




Which kind of sensors are required?


There must be sensors applied to the machine that can produce large amounts of data and provide information about the machine’s internal problems. In mechanical systems, these are mostly vibration sensors. The more context information is available, the better the software can work. Therefore, the following applies: the more sensors and data, the better.




What can I do if I don't have any sensors yet?


Companies whose machines don’t have suitable data-producing sensors can retrofit the required sensors. ai-omatic can support companies with such a retrofit.




With what frequency does the data have to be collected?


ai-omatic needs data with a frequency twice as high as what is to be monitored. This means: for a motor that rotates at 1000 revolutions, a sampling rate of 2kHz is required.

Since the required frequency is highly dependent on the specific application, it is difficult to define guidelines. Based on experience, however, it can be stated that at least one measurement per second must be available (1 Hz) so that the data is suitable for the software. High-frequency data, i.e. data in the range of 1 kHz and more, is more ideal.




In which format does ai-omatic need the historical data?


Historical data can be transferred in different formats, a widely used format is CSV.




What time period should the historical data set be from?


In general, there should be at least one month of historical data available. However, the amount of data required is heavily dependent on the frequency of the data and the type of machine. For machines with a constant, homogeneous behavior in normal operation, which generate data in the high-frequency range, data of less than a month is usually sufficient. In the case of data that is not high frequency in the Hertz range, this period increases to approx. 1-3 months.




If historical data is not yet available, how long does it take for a model to be trained?


Here, too, the amount of data required depends on the frequency and the type of machine. In general: for data in the kHz range, ai-omatic usually needs less than a month, for data in the Hz range approx. 1-3 months.




How does ai-omatic access the data?


Here, too, the amount of data required depends on the frequency and the type of machine. In general: for data in the kHz range, ai-omatic usually needs less than a month, for data in the Hz range approx. 1-3 months.




How much internet bandwidth do I need for the software to work?


The bandwidth required for sensor data streams depends on the amount and frequency of the data. In most cases a standard internet connection of 50 Mbit/s is sufficient.




Where is the software running?


The software runs on an ai-omatic cloud server operated by Hetzner. This is a cloud protected by firewalls with GDPR-compliant operation. The locations of the data centers are in Germany and are subject to strict European data protection regulations. In selected cases, ai-omatic also offers a more complex and cost-intensive on-premises solution, in which the software is integrated into the customer's infrastructure.




How does ai-omatic process my data and how is it used?


ai-omatic takes data protection very seriously. Therefore, data is only used in the context of project work with the customer and not beyond. As of January 2022, ai-omatic will also be certified with the international standard ISO/IEC 27001, which specifies the requirements for the establishment, implementation, maintenance and continuous improvement of a documented information security management system taking into account the context of an organization as well as requirements for the assessment and handling of information.




What are the costs?


ai-omatic aims at enabling companies to get started with predictive maintenance as easily and cheaply as possible. The workshop, in which the use-case and goals of the project are discussed, costs € 500, the quality check of the data costs 500 € and a one-time setup fee of 15,000 € is charged for the proof of concept. In the subsequent SaaS model, the price depends on the desired model: the essential model is €499 per month, the premium model is €699 per month and the custom model starts at € 899per month.




What exactly do I get from ai-omatic in the end?


The customer receives a user-friendly dashboard, that offers an overview of the vitality status of the machines. With this, failures can be predicted up to six hours in advance and thus maintenance costs can be drastically reduced. New machines or data sources can be added easily and intuitively by the customer and produce new results within a few minutes. Thanks to the web-based software, access is possible from anywhere.