PYTHIA Prevention

Live prediction and prevention of process disruptions

With PYTHIA Prevention, process disruptions, especially those in highly complex processes, can be significantly decreased.

PYTHIA Prevention creates a prediction model fully automated, which predicts the disruption based on a live data stream. Combined with your domain knowledge, you gain a solution that recommends the right counter-measures at the right time to prevent disruptions.


Applying PYTHIA Prevention into your process can bring significant benefits:

Understanding the reasons

Get a better understanding of the reasons for disruptive events and which kinds of different situations can cause them.

Quick and simple

Challenging and time-consuming tasks like data cleaning, data analysis, and model generation are done automatically by PYTHIA.

Useable by non-experts

Turn every operator into an expert: as soon as a potentially dangerous state is predicted, a counter-measure, defined by experts, is recommended to the operator.

Increase efficiency

Safe money by significantly decreasing disruptions and lower safety margins and maximizing throughput.

Proven value

The paper-making process is among the most complex processes. Avoiding sheet breaks there is as crucial as it is difficult. PYTHIA Prevention has proven its benefits there.

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.

How it works

PYTHIA Prevention is a combination of data analytics, prediction, and domain knowledge. Setting it up for your use case is done as a project following these steps:



You provide the raw, unfiltered history data of your process. The more data, the better. An in advance exclusion of individual signals or parts of the data because your domain knowledge tells you they seem not relevant for the target is neither necessary nor recommended. You then define a rule, how to identify the target disruption within this data, e.g., ‘the disruption occurs as long as sensor_318 is zero’. Also: The data conversion is performed.*



A fully automated root cause analysis on the history data is performed by PYTHIA with respect to the defined target disruption. PYTHIA also does a classification of the root causes since the same disruption might occur in different situations. Each of the found classes represents a set of relevant root cause indicators, including their importance.


Definition of counter-measures

You domain experts define suitable counter-measures for each of the classes. This is usually a straightforward process since they know exactly what to do as long as they know in which state the process is.


generation of a prediction model

PYTHIA generates a prediction model for the found classes. This model is designed to operate on a live data stream while watching all relevant parameters and predicting the imminent occurrence of one of the found situations.


Integration and deployment

The prediction model is implemented on-site (cloud-based or on-premise) and fed with the processes live data stream. The model now continually watches all relevant process parameters and informs the operator as soon as the process threatens to slip into a state that could result in a disruption. It then recommends suitable counter-measures to prevent the disruption.

*Your data needs to be converted into the required data format (see data format specification). You can either do this by yourself or use our data conversion service. You can choose to do so when filling out the registration form.

Talk to an expert!

Do you need someone to talk? We’re all ears.

* 30 minutes free


Time-series data specification.pdf
Tabular data specification.pdf

Want to understand, what’s behind PYTHIA Prevention?