Finished product movements are often constrained by laboratory certification turnaround time and the uncertainty introduced by measurement variability, contributing to logistic limitations and higher planning buffer. An alternative certification, or online certification, approach based on spectroscopy, for example, online Fourier Transform Infrared (FTIR), Raman, or Nuclear Magnetic Resonance (NMR), combined with chemometric models can provide fast, high-frequency, and high-precision property predictions that enable tighter control of blend targets, reduced quality giveaway, and more agile optimization cycles. This article outlines a practical framework to deploy spectroscopic analyzer systems and chemometric modeling for refinery product optimization while aligning with industry practices for performance-based qualification validation in ASTM D8340, model development and deployment practices in ASTM D8470, and multivariate model validations in ASTM D6122.

Refineries typically optimize blends across multiple property constraints while balancing economics, operability, and contractual specifications. In practice, the optimal target is not simply the specification limit; it also includes a buffer for measurement variability, sampling variability, and potential dispute risk at the compliance point.

When laboratory results arrive hours after a blend movement, operators often increase planning offsets to reduce the risk of re-blend or off-spec product. This creates quality giveaway and leaves margin in the blend. Many refineries have adopted online analyzer systems over the years, often integrated with blend optimizer software; however, conventional analyzers such as knock engines and volatility analyzers can be maintenance-intensive and are typically limited to one or a narrow set of properties. By contrast, a spectroscopic analyzer paired with a validated chemometric model can generate high-frequency, multi-property predictions in minutes-or even seconds-enabling faster optimization cycles, earlier drift detection, and lower target offsets while maintaining confidence in conformance decisions.

Alternative certification with spectroscopy is credible only when it is governed by the industry standard compliant qualification and validation processes. The three ASTM practices referenced here were published under ASTM D02.25 committees to provide guidance to users on how to build the processes for calibration and validation.

• ASTM D8340 – a performance-based approach to qualify a spectroscopic analyzer system for a defined use case, including the analyzer, sample interface, model, and ongoing controls.
• ASTM D8470 – a framework for developing and deploying spectroscopic/chemometric prediction models and establishing model robustness for the intended product and process space.
• ASTM D6122 – Validating the performance of multivariate spectroscopic analyzer systems (online, at-line, field, or laboratory), including leverage/outlier handling and ongoing verification strategies.

Together, these practices support a disciplined workflow: (1) define the certification intent and boundaries, (2) build and control the chemometric model, (3) qualify the analyzer system against performance criteria, and (4) validate and continually verify analyzer predictions for blend optimizations.

Spectroscopic analyzers infer product properties by measuring how a sample interacts with light across a wavelength range. In refinery blending and certification, the most common deployments are online analyzers installed on blend headers, rundown streams, or tank recirculation loops. A typical system includes a sample takeoff, conditioning (pressure/temperature control, filtration, phase management), an optical cell, the spectrometer, and software that converts spectra into predicted properties based on chemometric models.

Chemometrics uses multivariate statistics to relate spectral features to laboratory reference values (primary test method result, PTMR). The model “learns” the relationship between spectral variables, latent variables or factors, and the PTMRs (for example, octane, vapor pressure, or distillation points). Model quality depends on representative training data that spans the expected compositional and operating space-feedstock variability, seasonal gasoline changes, component swaps, and unit upsets.

Because spectroscopy is an indirect measurement method, strong governance is required around spectral pre-processing, instrument qualification validation, and outlier detection. Multivariate controls such as leverage statistics and spectral residual metrics help detect when a sample is outside the model’s domain (for example, unexpected blend components or abnormal volatility). These controls-paired with routine verification against lab results-are central to sustaining compliance with validation practices such as ASTM D6122 and to supporting a performance-based qualification approach such as ASTM D8340.

1. Define certification scope and risk assessment

Identify products (for example gasoline grades), properties, the point of compliance (tank, pipeline, custody transfer), dispute process, and the acceptable risk of re-blend or claims. Translate this into internal acceptance limits and optimization buffers (or blend target buffer).

2. Collect the primary test method dataset

Collect corresponding spectra and PTMRs across the expected compositional space. Ensure samples cover all grades, blend component variability, and abnormal operating conditions.

3. Develop chemometric models (ASTM D8470):

Select preprocessing, choose model type, and partition data for calibration and independent validation. Qualification validation to be completed as per ASTM D8470.

4. Validate the model and analyzer performance (ASTM D6122)

Chemometric model performance validations.

5. Qualify the analyzer system for the intended application (ASTM D8340)

Treat qualification as system-level: sample handling + instrument + model + operational controls. Establish performance-based acceptance criteria and demonstrate they are met under real refinery conditions.

6. Deploy into the optimization loop

Stream predicted properties to the blender/optimizer (or advanced process control) at high frequency. Use the fast update rate to run more frequent optimization cycles and reduce conservative offsets.

7. Manage prediction bias and drift (ASTM D6122 / D6708)

Implement routines to detect and correct systematic bias.

8. Continual validation and change management

Maintain ongoing verification with clear pass/fail criteria, handle probationary operation when data are limited, and control changes to instruments, sample systems, and models with requalification as needed.

High-frequency analyzer predictions shorten the feedback loop between a blend action and a quality response. Instead of waiting for infrequent lab results, the refinery can adjust blend ratios as properties approach constraints, which supports smaller planning buffers and reduced giveaway-especially for high-value constraints such as octane and vapor pressure.

A single spectroscopic analyzer can predict multiple properties through dedicated property models, allowing the optimizer to evaluate several constraints at once. This reduces the need to maintain multiple standalone analyzers. Because spectroscopic analysis is non-destructive, it also typically requires less maintenance than conventional analyzer technologies.

• RACI and ownership

Define responsibilities across Lab, Operations, Process Control, and Planning/Economics for model stewardship, analyzer maintenance, and certification decisions.

• Model and instrument configuration control

Maintain versioning for models, preprocessing steps, instrument settings, and sample-system configuration. Treat any material change as a managed change with defined revalidation or requalification steps.

• Validation dataset coverage

Ensure validation samples span the intended compositional space and include edge cases near specification limits where certification risk is highest.

• Outlier and “domain of applicability” rules

Establish objective criteria for rejecting or flagging predictions when leverage or residual statistics indicate the sample is outside the model domain; define operator actions when alarms occur.

• Bias and precision tracking

Trend analyzer-vs-lab differences over time, manage bias corrections with documented rules, and trigger investigations when drift exceeds limits.

• Continual validation cadence

Set verification frequency based on risk and variability (for example, higher frequency during seasonal transitions or after major component slate changes) and document pass/fail criteria and escalation paths.

• Audit-ready records

Keep evidence packages for qualification, validation, ongoing checks, and corrective actions-supporting the performance-based qualification philosophy referenced by ASTM D8340 and ASTM D6708.

1. Pilot (4-12 weeks)

Select one product grade and 1–3 critical properties; install or repurpose a spectroscopic analyzer; begin collecting matched spectra + PTMRs; build an initial model and define outlier rules.

2. Qualification and controlled use (8–16 weeks)

Execute validation per ASTM D6122 expectations; establish the analyzer-system qualification dossier per ASTM D8340; deploy predictions to operations dashboards and run in “shadow mode” alongside lab certification.

3. Optimization integration (4–12 weeks)

Feed validated predictions to the blend optimizer; reduce planning buffers gradually based on measured performance; implement alarms, fallbacks, and bias-management routines.

4. Scale and sustain (ongoing)

Expand to additional grades/properties; formalize continuous validation routines; institute periodic model refreshes to handle seasonal and feedstock shifts; measure value realization (giveaway reduction, re-blend rate, dispute rate, and analyzer service factor).

Alternative certification enabled by spectroscopic analyzer technology and chemometric modeling can shift refinery blending from a low-frequency, lab-lagged practice to a high-frequency, prediction-driven optimization loop. When implemented with disciplined model development, system qualification, and ongoing validation aligned with ASTM D8470, ASTM D8340, and ASTM D6122, refineries can reduce uncertainty, safely tighten blend targets, and improve economics-while maintaining confidence in quality conformance decisions.