Beyond the data deluge

In the quest to improve throughput, efficiency, safety and reliability of their operations, heavy industries are ramping up digital-transformation initiatives. Oil & gas, mining and metal companies are embarking on multiyear journeys to digitize operations by adding connectivity, automation and data analytics.

smart industry

Falkonry's Sachin Andhare


According to McKinsey & Company, most heavy-industry sectors are at the middle stages of digital maturity (Digital 2.0), relying on rule-based automation and distributed control systems. Others have made progress in digital maturity (Digital 3.0) and use collaborative robots, advanced process-control systems and real-time optimizers.

However, a few digital pioneers are applying artificial intelligence (AI) and machine learning (ML) to operational data (Digital 4.0), which enables predictive operations.

So what can we learn from these digital pioneers? And how can companies apply ML across industries at scale?

Data deluge

Manufacturing and process engineers need to review operational data across many sources such as ERP systems, MES, historians, computerized maintenance-management systems, etc. Due to the sheer volume and streaming nature of data, manual inspection and analysis is unsustainable. Often data from diverse systems comes in multiple units, formats and incompatible protocols. To get a better handle on data, companies have invested in building massive data warehouses, data marts and data lakes, but that has brought new challenges related to data governance and validation. Industrial-operations teams need a scalable, robust, and high-performance environment for operational data management, review, learning and analysis.

Industry challenges

Applying AI and ML techniques to operational data has been challenging for a variety of reasons. Most AI/ML platforms are designed for data-science professionals, not the operations teams. These platforms are difficult to learn and cumbersome to use. Moreover, the standard ML process is complicated and laborious, with long lead times of 6-9 months. It involves data preparation, blending, feature engineering, algorithm selection, model training, testing and deployment.

Once built and tested, data scientists hand over these custom models to operations. There is little collaboration with the operations staff and the majority of data-science initiatives fail to achieve their objectives. Also, the output of these models lacks transparency and does not give the subject matter experts (SMEs) in operations confidence to trust the underlying technology in a production environment. These SMEs trust their “intuition” or gut feeling over any AI program and, hence, the custom models rarely get deployed beyond pilots.

Predictive operations at scale

The solution is an ML system designed and built for industrial operations; a system that is usable by operations teams without requiring additional resources. A collaborative ML system allows operational experts to leverage their domain expertise by providing relevant condition facts or labels based on historical data. Then, by using unsupervised and semi-supervised ML algorithms, the system finds patterns in production data, identifies deviation (in quality as an example) and empowers SMEs to prevent recurrence.

This ML system must hide the complexities, automating feature engineering and other onerous steps of ML process. To make ML transparent, it should quantify prediction and provide explanation and confidence metrics with details into model output. Explainable AI gives insight into predictions, allows users to understand signal correlation, and enables SMEs to understand which process variables affect outcomes for root cause analysis.

Industrial operation presents multi-dimensional problems, whether you are optimizing efficiency, preventing quality defects or reducing asset downtime. A domain-agnostic, prepackaged ML system offers analytical flexibility to address multiple use cases and dramatically improves the life of operational experts, as they don’t have to learn new tools for each new project.

This is the preferred approach among digital pioneers—it saves time, allows repeatable ML process and delivers superior ROI.

A collaborative ML system that automatically identifies patterns among hundreds of variables, delivers actionable insights and generates SME trust will enable predictive operations. And empowering operational SMEs is the way to achieve a Digital 4.0 maturity level in any industry at scale.                    

Sachin Andhare is the global marketing director with Falkonry.

Want more on predictive operations? Find our library of features here.