_

Seeq Add-ons

Our offer includes custom packages developed on Seeq to perform specific calculations. These toolsare intended to help the user in performing complex analyses and understand their results. Utilization of both visualization techniques and mathematical capabilities allow such tools to help the user in insightful analysis performed on Seeq’s data platform.

PRINCIPAL COMPONENT ANALYSIS

Introduction

Principal Component Analysis (PCA) is a powerful statistical technique widely applied in industrial and operational fields for anomaly detection and root cause analysis. By reducing the dimensionality of complex datasets, PCA identifies patterns and relationships that may not be immediately apparent, enabling the detection of deviations or anomalies in system behaviour. These anomalies often indicate potential faults, inefficiencies, or abnormal operating conditions. Moreover, PCA’s ability to highlight the most significant contributing variables aids in pinpointing root causes, facilitating targeted interventions. Its application is particularly valuable in environments with high-dimensional data, such as manufacturing, energy systems, and process control, where operational reliability and efficiency are critical.

Installation

The following are the installation requirements and procedure.

Prerequisites

If you want to install seeq-ITV-visual as a Seeq Add-on Tool, you will need:

  • Seeq Data Lab (>= R52.1.5, >=R53.0.2, or >=R54);
  • Seeq module whose version matches the Seeq server version;
  • Seeq administrator access
  • Enable Add-on Tools in the Seeq server.

Installation

  1. 1. Create a new Seeq Data Lab project and open the Terminal window.
  2. 2. Run pip install BUILDER_ADDON_PCA

Utilization and performance

This SEEQ add-on connects automatically to the worksheet from which the add-on was launched. The principle is to apply the PCA formulae to make a comparative analysis of the different parameter influences within a closed system. The main contributions will be those of the variables that most influence the deviation from the behaviour that PCA understand as ‘normal’.

A train condition will be used to train the model on the correct behaviour. This period is expected to have a minimal percentage of anomalies.

A test condition will be used as the period where the PCA must calculate the different parameter influences within the time range. It is possible to give the Add-on a condition with multiple time capsules. In this case the PCA will be applied on each time capsule.

A phase condition (optional) allows the add-on to be configured on a different setting:

  • Test condition defines the macro time capsules where PCA must be calculated.
  • Phase condition defines the relevant time periods where signals must be used for Principal Component Analysis.

(e.g. If you want to calculate the Principal Components for all the start-up periods in the last month your test condition will be the last month and the phase condition will be the start-up periods)

Following a brief description of all the parameters used in the calculation.
Parameter Description
Time Interpolation Time Interpolation Defines the sampling frequency of data from Worbench
Data Reduction Defines the amount of variables reduction (0% no reduction)
Save Output If enabled all outputs are written on a new Seeq Workbench

Examples

We used internal data to store variables related to a specific asset. Inside we defined a Train condition and a Test Condition to use in our Principal Component Analysis

From the PCA Add-on fulfill each field, then Submit your parameters.  It is possible to select all available conditions on the worksheet inside of the Train Condition Selection Box and the Test Condition Selection Box.

During the calculation the model will use the training period data to teach itself the normal condition settings for the system of variables defined in the workbench. Once the analysis is performed  the add-on will show how each variable affects the system’s deviation from the normal condition. This model is particularlu usefull to detect anomalies in stationary scenarios

The graph illustrates how each variable affects the system’s deviation from the  for a specific time capsule. The title provides the reference for the time capsule, including the start time of the corresponding period. Clicking on the “Anomaly Detection” icon on the bottom-right corner it is possible to visualize the deviation of the system from the normal condition for each observation.

Using the Pick Event bar below the graph it is possible to go through all the time capsules loaded onto the add-on for the Test Condition.

Contacts

For more information, feel free to contact us at software@e-matica.it

Scroll to Top