Expectator

The Expectator is a software suite for the automated plausibilisation and the detection of errors in numeric data. Especially it is designed for the identification of anomalies and incidents within sensor signals and time series data of arbitrary types.

 

  • Purpose
  • Features
  • Architecture

Normally there is no anomaly or incident same than another one. Therefore we do not believe in an overall one-for-all algorithm for incident and anomaly detection. Instead the Expectator follows the philosophy to define numerous simple and pragmatic criteria for the description of the expected regular characteristics and shape of sensor signals and combine these criteria to powerful committees with much less wrong classifications (false positives). For this a lot of methods with different complexity and functionality are available.

In this way the Expectator is a tool for the quick and pragmatic development, formulation, definition, calibration, administration, evaluation, and adaptation of so-called expectations to the content of arbitrary sensor signals. For this we distinguish between the generic tool for editing and administrating the expectations (offline) and specialized tools for executing and evaluating the expectations in the according embedded target environment (online). Latter normally is customized for the corresponding application domain and environment.

 

  • Definition, calibration and adaption of various methods for the formulation of expectations to normal signals
    • checking the values of simple signal statistics from single and combined signals
    • checking if signals lay within given bands and domains
    • checking if signals cross certain intervals or timely and spatial domains
    • checking the distance to given reference signals with arbitrary different signal metrics
    • checking the distance between the given signals
    • usage of Self Organizing Maps for the identification of anomalies
    • checking the difference between real measurements and predictions by arbitrary predictive algorithms
    • checking signal characteristics in the frequency domain for transient and periodic signals
    • … 
  • Customer and application specific adoption and integration of the evaluation for arbitrary target platforms
  • Expandable with arbitrary expectation methods for data plausibilisation and calibration of the expert system

The Expectator is a MATLAB-Toolbox, which is based on the Signal Structure Toolbox of ANDATA, which is distributed open in common with Expectator and Stipulator.

All functions of the Expectator can be controlled by a Command Line Interface, allowing to compose workflows in application specific scripts and apps for an easy integration into custom environments.