Knowlestry GmbH
Vimystraße 1e • 85354 Freising
E-Mail: ia.yrtselwonk@ofni

Our hours of operation
Monday - Friday 9 am to 6 pm

“Shorten today's lengthy
and expensive AI development.”

Knowlestry automates the interaction between your domain knolwedge, your data and AI.
With Knowlestry you are maximizing value creation from data.

Our clients' alternative approach to high quality data-driven development involves 4 steps:

  • Describe Use Case
  • Master Domain Aspects
  • Obtain & Manage Analytics Results
  • Refine Process

Our Knowlestry Data Science Team is highly motivated for outstanding projects together with industry partners.

Start your transformation here

Set up and successfully use Knowlestry technology on your own. Get support and updates directly from our core development team.

Continue your evolution here

Take a look at some application examples.

Virtual commissioning of electric motors

An electric motor manufacturer in Bavaria, Germany, offers virtual commissioning as a digital service to its customers. Knowlestry technology was used to create an AI model to classify different operating states of electric motors.

Requirements:

  1. Classification based on electric current, rotational speed, and motor temperature
  2. Applicable to different motor sizes and environmental conditions
  3. Address data quality issues

Results:

  • Automated handling of domain-specific aspects:
    • missing or wrong values in the records
    • delays between power change and temperature change in processing and selecting training data
    • Influence of ambient conditions such as temperature, heat dissipation
  • Availability of domain-specific knowledge as semantic embedding for automatic reusage in other use cases
  • AI Model to predict the operation state
  • Feedback of data quality and completeness of data

Predicting production downtimes

An electronics manufacturer in Bavaria, Germany, wants to reduce the cost of production downtime. Using Knowlestry technology, an AI model was developed to predict and classify downtime.

Requirements:

  1. Predictions based on an event log of production steps and electric power data
  2. Detection of downtimes and classification of its cause
  3. Applicable across multiple production processes

Results:

  • Automated handling of domain-specific aspects:
    • Synchronization of data from multiple sources based on timestamps
    • Interpolation of missing values
  • Availability of domain-specific knowledge as semantic embedding for automatic reuse in other use cases
  • AI model to predict and classify the downtimes
  • Feedback on the efficiency of the individual data sources, giving a recommendation on what data to capture

Condition monitoring of frictional brakes

A Bavarian manufacturer of friction brakes provides digital services with its products. Knowlestry technology supported the development of an AI model to detect whether the brake is applied or released, based solely on electrical voltage and current measurements.

Requirements:

  1. State detection based on voltage and current
  2. Handling of environmental influences such as: (A) supply voltage fluctuations (B) non-relevant AC components in the supply voltage
  3. Balancing the risks of false positives and false negatives

Results:

  • Automatic handling of domain-specific aspects:
    • Filter voltage/current by removing irrelevant parts such as AC component
    • Selection of training data to cover all important scenarios, e.g., mechanically locked brake, fluctuating supply voltage
  • High-accuracy AI model to classify the applied and released states
  • Interpolation of missing values
  • Availability of domain-specific knowledge as semantic embedding for automatic reuse in other use cases

Audit in groundwater management

A groundwater management service provider wants to reduce the manual effort required to monitor and report changes in the groundwater levels. The Knowlestry technology facilitates the development of an AI model to automate this task.

Requirements:

  1. Detection of special events and changes in various parameters, e.g., groundwater levels and water flows
  2. Reported explanation of the event in text form
  3. Prediction based on a large amount of previous data, but with many inconsistencies in previous reporting

Results:

  • Availability of domain-specific knowledge as semantic embedding for automatic reusage in other use cases
  • Based on this knowledge and previous reports, a structured, unified and standardized system for the text-based reports
  • AI models for the detection of relevant events and their explanation in text form
  • Automatic handling of domain-specific aspects:
    • Mutual influence of geographically close machines
    • Influences on groundwater levels outside the system
    • Correcting and handling faulty and non-plausible data