Client

Metals/process industry plant

Industry

Metals & mining/process manufacturing

Primary goal

Process optimisation and quality control

Technologies

Machine learning multi layer perceptrons with real-time analytics, IoT data integration

In process industries, maintaining optimal operating conditions is critical for throughput, quality, and stability and to maintain the trade offs within those. Traditionally, key parameters such product compositions are measured manually through lab tests every two hours, leading to delays in corrective actions.

KPMG in India developed a machine learning-based soft sensor to predict quality in real time, enabling proactive operational adjustments.

The challenge

Quality is a critical parameter in oxidative furnace processes, directly impacting productivity and downstream quality.

However, reliance on periodic lab testing created delays in detecting deviations, resulting in suboptimal process control, variability in process parameters, and lower efficiency.

The opportunity

KPMG in India developed a real-time predictive model using machine learning multi layer perceptrons techniques such as XGBoost and RNNs with LSTM to estimate SS ratios continuously.

The model integrates multiple data sources, including high-frequency process data, composition data, and operational parameters, to generate accurate predictions.

These predictions enable operators to proactively adjust parameters such as airflow, feed rate, and temperature.

KPMG in India’s approach included:

Deep Domain Understanding

Process Chemistry, Reaction Kinetics, Mass and Heat balances
Data integration

Combining LIMS, OSI PI, and real-time sensor data
Stack Build

OPC UA and API layer build, building time series database
Advanced modelling

Using ML algorithms for predictive accuracy

Feature engineering

Applying PCA and domain-driven variables
Real-time deployment

Integration with operational dashboards
Action enablement

SOPs for operational adjustments based on predictions

The impact

The solution enabled a shift from reactive to predictive process control:

monitor

Real-time monitoring and prediction of quality of exit products of the oxidative furnaces

insights

Improved process stability and consistency

arrow_upward

Increased throughput and production quality

insights

Faster decision-making through actionable insights

science

Reduced dependency on delayed laboratory measurements

How we make the difference

KPMG in India leverages advanced analytics and real-time data integration to enable intelligent process control in industrial environments. Our solutions help organisations optimise performance, improve quality, and achieve operational excellence through AI-driven insights.