PYTHIA Virtual Sensors
Fast and Automated Generation of Virtual Sensors
With PYTHIA Virtual Sensors, creating virtual sensors becomes quick and straightforward. No data science expertise is required. It takes all necessary steps unsupervised – from data cleaning and synchronization to model generation.
PYTHIA Virtual Sensors bring a variety of great advantages:
Unsupervised model creation
Every software sensor needs a model to calculate the sensor value from other parameters. This model is being created without human supervision; all it needs is a relatively small amount of learning data.
No manual data cleaning
Tasks like data synchronization and cleaning, removing anomalies, and adding sensor values used to take up a lot of time. AI Sensors automates all these manual data preparation tasks.
No data science needed
All of the work that conventionally would be done by a data scientist, is done automatically. Therefore, no data science expertise is needed for generating a Virtual Sensor.
Even in complex scenarios with data from thousands of physical sensors, a virtual sensors can be generated within one hour.
The virtual sensor can be used in closed-loop and real-time scenarios since its able to yield measurements within milliseconds. It can also easily be deployed on-edge.
Based on PYTHIA
PYTHIA works unsupervised and provides faster results with lesser and unprocessed data while handling thousands of parameters and uncovering root cause indicators.
Applications for Virtual Sensors
The field of application and use of Virtual Sensors is huge:
Replacing real sensors
Sometimes physical sensors are too expensive or are located in an environment where they are prone to failure or have a short life expectancy.
Measure impossible things
Determine abstract parameters like quality or lifespan, which cannot be measured by any physical methods.
Reducing measurement errors of real measured values based on correlations with other physical measured values.
Less lab measurements
Create virtual sensors for live measuring parameters that otherwise can only be determined in the lab with a time delay.
Measure non-measurable variables such as the temperature inside a reactor with a soft sensor that has been trained using simulation data.
Increase number of sensors
Increase the number of sensors to monitor your process by combining virtual and physical sensors.
Virtual Sensor Building Service
Use this service to quickly create virtual sensors.
Fill out the self service registration form with your personal and order details.
We send you an confirmation eMail with the offer and an NDA.
You receive instructions how to upload your data and how to define your target.
You can either do this on your own or use the data conversion service.*
We cerate the sensor according to your defined target.
You receive a link to the sensor and final report for download.
PYTHIA Virtual Sensors
virtual sensor building Service
*In order to use this service, your data needs to be converted into the required data format (see data format specification). As an option, you can make use of our data conversion service. You can choose to do so when filling out the registration
Virtual Sensor Building App
The AI Sensors app was designed to let you quick and simple create your own Software Sensors based on data you converted and uploaded to the app.
A free demo of AI Sensors is now available. See below.
Create a Virtual Sensor step by step
Creating a Software Sensor step by step.
Step 1: Prepare dataset (see specifications in download section)
Step 2: Upload dataset
Step 3: Check uploaded dataset
Step 4: Creating sensor, select dataset
Step 5: Creating sensor, define sensor target signal.
Step 6: Creating sensor, choose parameters.
Step 7: Create sensor.
Step 8: Wait until sensor was created.
Step 9: Check quality of the sensor.
Step 10: Download the sensor model.
Step 11: Deploy the sensor.
Talk to an expert!
Are you not sure what you need? We’re all ears.
* first 30 minutes are free
Simple water pipe simulation
Simple simulation example of a water pipe, where the water input and the switch vary randomly. Y measures the output volume and is replaced by a Software Sensor.
x1: Amount of water (PerlinNoise between 0l and 10l)
x2: valve with state in percent (0% to 100%)
y: Amount of water (calculated value in l)
Water is flowing into a pipe from the left, passes a switch and flows out through two outputs. The water quantity flowing in as well as the switch position change randomly. There are two time lags between X1 and X2 and between x2 and Y. This means, that however X1 and X2 change, the outgoing flow at Y will see those changes with a time delay.
The sensor for the water flow at Y was now replaced by a SoftwareSensor, that only sees the current values of X1 and X2, predicting the water outcome at Y.