Data Explorer Overview
Data Explorer helps you prepare data for new discovery and identify relationships with statistics or model analysis by allowing you to:
- Graphically interact with data relevant to manufacturing analytics.
- Learn new process relationships.
- Identify and eliminate data issues and enrich data to leverage existing knowledge.
The first step in Analytics is gathering data, aligning different data sources (by time usually), and preparing it for learning about a system/problem/process. While Machine Learning (ML) algorithms allow you to develop useful models to improve manufacturing, the raw data may have the following issues:
- Not enough aligned data from diverse manufacturing systems.
- The data has problems or significant periods that are irrelevant to the task at hand (for example, out-of-service times, upset conditions, relevant grade, etc.).
- The data has pure data management issues (for example, excessive compression or noise).
Data Explorer enables you to visualize and graphically interact with your data in order to observe and detect issues. Using the array of tools provided by Data Explorer, you can then perform the following tasks:
- Visualize past data for information discovery.
- Clean data graphically.
- Perform a quick and flexible time merge.
- Prepare statistical summaries.
- View correlation and ranking, including time.
- Transform data for enrichment without programming.
- Filter and align data for useful machine learning and
- Generate streaming calculations in preparation and consistent with the ML data
The ultimate deliverable is useful correlations that can be identified by an informed user to detect new process/system influences and causes. You may also export the determined useful data in formats ready for many machine learning packages. In this version, this functionality can also be integrated with manufacturing focused ML application development platform for an even simpler workflow.
With the new Soft Sensor
®
capability Data Explorer can now develop or target causally correct models that support explain-ability including influence ranking and model response curves. Additional system understanding is available from modeling results. Finally, users can now not only export prepared datasets with streaming data prep calculation scripts but export real-time Soft Sensor calculation scripts to calculate prediction results from these models.Provide Feedback