NIR Gasoline Blend Models for Accurate Octane Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fuel blend models are labor-intensive and prone to deterioration, leading to excess octane production and reduced profitability due to the need for numerous models tailored to specific fuel grades and seasonal changes, which increases maintenance costs and decreases model performance.
Innovation Solution
Development of generalized blend models using near-infrared spectroscopy and machine learning algorithms to predict octane numbers, incorporating spectral data from multiple sources and refining processes to create models that are more broadly applicable and maintain consistent performance across various gasoline blends.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple fuel grade-specific blend models are maintained to minimize excess octane, then blending optimization is improved, but model maintenance labor intensity increases
Solution Approach 1:
The patent applies universality by developing a single blend model that can handle multiple fuel grades and seasonal variations rather than requiring separate models for each grade. The model uses generalized spectral features and machine learning algorithms that work across different fuel types, eliminating the need for grade-specific models while maintaining optimization capability.
Solution Approach 2:
The patent merges multiple fuel grade-specific models into one unified blend model. It combines spectral data from different fuel grades and seasonal variations into a single database, training a machine learning algorithm to handle all cases simultaneously. This consolidation reduces the number of models from multiple to one, significantly reducing maintenance labor.
2Manufacturing precision
If blend models are tailored to specific seasons and fuel grades, then blending precision is improved, but device complexity increases
Solution Approach 1:
The patent creates a universal blend model that handles multiple fuel grades and seasonal variations through a single machine learning algorithm. The model is trained on diverse spectral data covering different grades and seasons, enabling it to accurately predict octane numbers across all scenarios without requiring separate specialized models for each.
Solution Approach 2:
The patent uses parameter changes by dynamically adjusting the model's predictive parameters based on input spectral data characteristics. The machine learning algorithm automatically adapts to different fuel grades and seasonal conditions by analyzing the spectral features, eliminating the need for manual model selection or switching between multiple fixed models.
3Measurement precision
If numerous blend models are developed for slight feed composition changes, then blending accuracy is improved, but loss of time increases
Solution Approach 1:
The patent applies parameter changes by using a machine learning algorithm that automatically adjusts its internal parameters based on the spectral data input. Rather than developing new models for each composition change, the single model adapts its predictive parameters dynamically, maintaining high accuracy while eliminating the time-consuming process of model redevelopment.
Solution Approach 2:
The patent implements self-service by enabling the blend model to automatically adapt to new feed compositions and seasonal variations without requiring external intervention to create new models. The machine learning algorithm learns from the spectral data and adjusts its predictions autonomously, eliminating the need for manual model development and reducing time losses.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The generalized models simplify maintenance, reduce labor costs, and enhance predictive accuracy, ensuring compliance with government specifications while minimizing excess octane production and maintaining profit margins.
Implementation Method 1
analyzing a first collection of liquid hydrocarbon samples comprising multiple finished gasolines by near-infrared spectroscopy to produce a first spectral database
Data Source
AI summary
Blending a finished gasoline using a gasoline blending model that is derived from correlations between empirically measured octane numbers and spectral features identified in near-infrared (NIR) spectral data for a group of gasolines and gasoline subcomponents. The correlations are incorporated into generalized blend models for motor octane number and road octane number, which are incorporated into programing executed by a controller that controls the volumetric blend ratio of one or more neat gasolines and/or gasoline sub-components to produce a finished gasoline. In some embodiments, the NIR spectral data utilized for developing the model is contributed by analysis of multiple subsets of gasolines and gasoline subcomponents, where each subset is analyzed by a different NIR spectrometer.