Smart Manufacturing- Steel Refractory using OCI Data Lakehouse

August 12, 2022 | 6 minute read
Siddesh C Prabhu Dev Ujjni
Senior Cloud Engineer
Text Size 100%:

The technical solution is based on the Oracle Lakehouse: the one platform to store, manage, enrich, and analyze all kinds of data. In this example, we are importing data feeds from Equipment and Products systems (API interfaces), running them through data quality and transformation pipelines to create a curated lake. The Oracle Lakehouse platform is the one place to ingest and manage all aspects of data in one location, and to keep it well-organized. From structured to non-structured, from relational to videos to tweets to geospatial, all of the formats are kept in one place for secure and organized analysis.

While all cloud vendors have a form of datalake, Oracle’s is the only one where you can freely combine the different types of data into one toolset. You don’t need to move data between different types of databases and you can use relational tools to query non-structured data or data science tools to do machine learning in-place in the lake. What’s more, the entire lakehouse shares one security model, one location, and one management approach for easier and more efficient operations.

Manufacturing Lakehouse

 

 

Specific components we use in the demo:

  • OAC (Oracle Analytics Cloud) is a one stop shop for all your reporting and business intelligence needs.
  • ADW (Autonomous Data Warehouse), the relational component of the Lakehouse for managing all the structured data
  • Object Store, the unstructured data component of the Lakehouse for managing all the non-structured data
  • Integration & transformation pipeline from external systems like Oracle Manufacturing: Data Flow Service, Data Catalog
  • DataSafe for identifying and managing confidential and PII data and its safety
  • IDCS and OCI security components for managing the authorization and security

If you observe the dashboard, it’s a well-organized and smartly categorized for your analytical needs by business area. Which gives a simplified view to all your business users starting from C-level executives all the way down to operational users for day-to-day work.

Under the Hood:

Solution - Under the hood

 

 

 

Discover: Here we are capturing the structured and unstructured data from different sources such as Batch Steel Refractory data and data from events and sensors.

Ingest: The data is ingested from a wide variety of source systems using OCI integration Services and passed over to the Data Lake object storage which is the landing Zone of the Lakehouse.

Transform and Curate: Further the data is cleansed, transformed, and enriched using the OCI Data Flow service, which is fully managed Apache Spark service, that performs processing tasks on extremely large datasets without infrastructure to deploy or manage. We move data from curated Zone to the final storage layer of Lakehouse which is a combination of ObjectStore and ADW. It helps democratize data across the organization.

Analyze, Learn & Predict: The data is now used to analyze and provide business insights through centralized OAC dashboards and open source Grafana on OCI. Further we extend our analysis using Oracle’s ML stack to learn & predict the outcomes.

Measure & Act: The Lakehouse platform can now be integrated with Custom mobile apps, Opensource tech stacks, hybrid platforms.

 

Solution Summary

  1. Steel Refractory - Executive Dashboard

The big picture

In this demo, an executive uses a dashboard of steel manufacturing refractory analysis to identify and isolate a distinct upward trend in manufacturing costs.

Note that while these look-like simple graphs of clean data, this data is rarely available in a simple format in a single place. While the business users mainly see the outcome, the actual process for getting here is complex and in the past, most organizations have been able to do this as strongly manual process maybe once or twice a year. Our solution enables you to analyze all this and more on an on-going basis.

To make the most sense, instead of relying on static classifications, we are letting AI find several of the categories and help with the modeling.

 

Executive Dashboard Demo

 

Exec Dashboard

Through the executive dashboard, plant manager is able to view how the different machine parameters are working. From the above dashboard the manager is able to view the parameters like gunning, Main Cabinet- Water, Main Cabinet – Hydraulic Oil, Terminator Cabinet but we can always add more features depending on the requirements of the customers.

 

  1. Visualize gunning material: We see from the graph on the top left how the gunning material fares and how much gunning material is produced per day across the week based on the datasets
  2. Visualize Terminator Cabinet: We see from the graph on the top right what’s happening from within the terminator cabinet in terms of the Lance Temperature and Oil pressure in bar.
  3. Visualize Main Cabinet – Water: The graph on the bottom left shows the temperature and pressure variations in the main cabinet due to water during steel production across different days.
  4. Visualize Main Cabinet – Hydraulic Oil: The graph on the bottom right shows temperature and pressure variations in the main cabinet due to hydraulic oil during steel production across different days.

Through all these visualizations, the plant manager is able to understand how the refractory is fairing during steel manufacturing across multiple days and multiple machine parameters.

 

Exec Dashboard-2

 

We have heard from multiple customers that the main cause of concern for the steel refractories are is costs that is endured if there is any maintenance needed and if there is no way to plan the maintenance then there is a lot of adhoc costs resulting from it which needs to be avoided.

 

Let’s see how the cost fares during the production of steel in the refractory.

 

Grafana-1

Grafana-2

 

 

 

Siddesh C Prabhu Dev Ujjni

Senior Cloud Engineer


Previous Post

‘Abracadabra’ and other AI myths

April Hawthorn | 4 min read

Next Post


Using Jupyter on OCI Data Science to analyze HPC results

Joe Hahn | 7 min read