Advanced photonics and AI in on-line process analysis: from measurement to operational insight

Across the process industries, on-line analytical instrumentation is undergoing a quiet but meaningful transformation. 

Historically, process analysers were deployed primarily to provide numerical measurements - concentrations, compositions, or physical properties - used for monitoring, control, or regulatory compliance. 

Increasingly, however, measurement is expected to serve a broader role: enabling real-time operational insight and supporting faster, more informed decision-making.

This shift is particularly evident in sectors such as refining, petrochemicals, and emerging hydrogen production, where tighter operating margins and heightened safety requirements place greater demands on measurement systems. 

In these environments, small deviations in process conditions can have disproportionate effects on safety, efficiency and product quality.

Moving beyond extractive sampling

A longstanding challenge in on-line analysis has been reliance on extractive sampling systems. 

While extractive approaches remain common, they can introduce delays, sample conditioning artefacts, and additional maintenance complexity. 

These limitations become more pronounced in high-pressure or fast-changing processes, where timely and representative measurements are critical.

As a result, there has been renewed interest in direct and in-situ measurement techniques. 

Optical and photonics-based methods - including absorption, fluorescence, and spectroscopic approaches - allow measurements to be performed directly at process pressure and temperature. 

By reducing dependence on sample handling, such techniques can improve response time and measurement robustness, particularly in chemically aggressive or hazardous environments.

For applications involving reactive gases or variable feedstocks, continuous, real-time data is increasingly viewed as a prerequisite rather than an enhancement.

From data to understanding

Advances in measurement capability have been accompanied by rapid growth in the volume and frequency of analytical data. 

Modern analysers can generate continuous data streams, but without appropriate interpretation, this data may add complexity rather than clarity.

To address this, advanced analytics and artificial intelligence are being applied to process analysis. 

Instead of treating analyser outputs as isolated values, analytical data is increasingly evaluated in combination with other process variables such as temperature, pressure and flow. 

This integrated approach enables trend recognition, anomaly detection and early identification of process deviations.

In refining applications, for example, real-time compositional data can be correlated with distillation behaviour to support faster responses to feedstock variability. 

In hydrogen production and handling, continuous monitoring of trace components can assist in identifying unsafe conditions before alarm thresholds are reached. 

Importantly, these methods are intended to complement, rather than replace, conventional control strategies.

Trust and measurement integrity

Despite the promise of advanced analytics, adoption ultimately depends on confidence in the underlying measurements. 

Process engineers remain cautious about relying on algorithmic outputs unless the physical measurement itself is demonstrably stable, transparent and well understood.

This places renewed emphasis on analyser fundamentals: sensor stability, calibration practices, diagnostics, and long-term drift performance. 

In many cases, the effectiveness of data-driven tools depends less on algorithm sophistication and more on the reliability and consistency of the measurement architecture.

Consequently, analyser suppliers and system integrators are increasingly expected to deliver complete measurement solutions, encompassing installation, validation, diagnostics and integration with plant automation systems.

Integration with automation and safety layers

Another notable trend is closer integration between analytical instrumentation and plant control architectures. 

Rather than operating as stand-alone devices, analysers are now expected to communicate seamlessly with distributed control systems and, in some cases, safety-instrumented systems.

This integration is particularly relevant in processes involving combustible or reactive gases, where analytical data may contribute to safety-related decision-making. 

In such contexts, the distinction between process monitoring and safety monitoring becomes less clear, and analysers must satisfy both operational and regulatory expectations.

Industry adoption

Within this broader landscape, instrumentation suppliers and engineering firms are aligning their offerings with these evolving requirements. 

UK-based Modcon Systems Ltd., for example, operates in the field of on-line process analysis, supplying analytical systems used in energy and process-industry applications. 

Like others in the sector, its work reflects a wider industry shift toward combining direct measurement technologies with data-driven interpretation to support operational decision-making.

Such examples illustrate a broader transition in which analytical instrumentation is positioned as an enabling layer between physical processes and digital optimisation tools, rather than as a passive measurement endpoint.

Looking ahead

As energy systems diversify and process conditions become less predictable, the demand for real-time, context-aware analytical insight is expected to grow. 

Analyser selection criteria are expanding beyond accuracy and repeatability to include diagnostics, integration capability and data usability.

While discussions often focus on digitalisation and artificial intelligence, the foundation of effective process analysis remains grounded in measurement physics and engineering discipline. 

Reliable decisions continue to depend on reliable data.

Gregory Shahnovsky is a process control and optimisation engineer with more than 30 years of experience in industrial process analysis and optimisation.

His work has focused on refinery, petrochemical, and energy-related applications, including real-time measurement, process safety and digital optimisation. 

He has authored numerous industry publications on process analysis, hydrogen safety and refinery optimisation.

He has been with MODCON since 1990 and currently serves as President and CEO of the MODCON Group.