One of the world’s largest natural resource mining and processing companies had its data lying in multiple siloed data repositories, files and databases. Its executives spent long hours extracting the data and preparing reports for the management, slowing down its routine decision making and making the process cumbersome.

Tredence developed a centralized enterprise Lakehouse which routinely synced data from multiple internal sources, 3rd party data sources and public databases.

Further, we built a Text-to-SQL chatbot connected to Databricks, capable of understanding domain-specific data, converting natural language queries into SQL for insight generation, ensuring secure and transparent responses, and using human feedback for continuous learning and improvement.

The cement manufacturer sought to improve product quality by changing its composition and production parameters based on its compressive strength test results. The quality managers tested cement samples from a batch at set intervals of 3, 7 and 28 days after curing. These elongated periods delayed corrective actions and increased the risk of batch rejections.

Tredence collaborated with the manufacturer to develop short- and long-term forecasting models using statistical and machine learning techniques to predict cement strength and reduce their reduced reliance on delayed 28-day strength tests.

Our early-age strength predictions, allowing quicker interventions to prevent defective batches and changes in composition and the clinker factor.

A leading aluminum product manufacturer suffered ongoing raw material quality issues that harmed production processes by creating excessive quality defects, work-in-process scrap, and equipment shutdowns.

The company wanted to predict the optimal machine settings to improve product quality.

Tredence built a system that deployed edge-based IoT services to extract product and machine data from an on-premises historian. Next, we developed real-time historian data ingestion pipelines and quality batch data ingestion pipelines from edge to cloud. We then developed and validated predictive machine learning models on Azure Databricks. Finally, we deployed predictive models at the plant edge to create low-latency insights and integrated them into a human-machine interface.

A Fortune 500 company that provides safety, security and sustainability solutions to aerospace, building automation and industrial automation companies face a challenge in helping the first responders in prioritizing the alerts.

The basic alert system could not provide information on the severity and the nature of the fire incidents. There would be multiple reports of the same incident coming from different sources and at times there would be false reports as well. All of this would result in inefficient rescue operations and resource and personnel losses.

Tredence partnered with the company to ingest the alert data and combine it with the historical data pertaining to fire incidents, their severity, location and other parameters to build a real-time fire system health and status monitoring user dashboards. The dashboard could identify unique incidents, suggestive probable causes and the likely severity which allowed the first responders to manage the response accordingly. The solution would discern false reports from the actual incidents conserving crucial manpower.