Naphtha Splitter Prediction Model for Yield Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chemical process simulators lack data to accurately predict product properties of naphtha splitter columns, making it difficult to optimize operations and improve aromatic yield in naphtha splitting units.
Innovation Solution
A method is developed to train a prediction model, such as a random forest model, using past operating conditions and product properties data to predict output variables like LSR D95, LSR D90, HSR D05, and HSR C6 paraffin, allowing for continuous prediction and operational control of naphtha splitter columns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If chemical process simulators are used to predict product properties of naphtha splitter columns, then operational control can be performed, but the prediction accuracy is insufficient due to lack of property data
Solution Approach 1:
A machine learning prediction model serves as an intermediary between available operating condition data and the required product property predictions. The model is trained on historical data and acts as a mediator that translates operational parameters into accurate product property predictions, bridging the information gap that chemical process simulators cannot fill
Solution Approach 2:
Instead of relying on physical measurement data that is only available once daily, the invention creates a virtual copy of the product properties through machine learning predictions. This digital replica provides continuous prediction values that mirror what actual measurements would show, enabling real-time operational control without waiting for physical samples
2Productivity
If product properties are measured once a day according to related art, then operational data is obtained, but continuous prediction and real-time operational control are not achievable
Solution Approach 1:
The machine learning model is trained in advance on historical operating condition data and corresponding product properties. This preliminary training action enables the model to make instant predictions without requiring new measurements, effectively preparing the system ahead of time for real-time prediction needs
Solution Approach 2:
The prediction model provides continuous prediction values for product properties, transforming the discontinuous once-daily measurement process into a continuous data stream. This enables uninterrupted operational control and eliminates the time losses associated with waiting for periodic measurements
3Manufacturing precision
If naphtha splitter column is operated based on operator's experience, then operations can be maintained, but optimization of aromatic yield and precision control are limited
Solution Approach 1:
The prediction model provides continuous feedback on product properties based on current operating conditions. This feedback loop enables operators to see the predicted impact of operational changes before implementing them, allowing for precise optimization of aromatic yield while maintaining a relatively simple operational framework
Solution Approach 2:
The invention replaces the reliance on human operator experience and judgment with an automated machine learning prediction system. This substitution transforms subjective, experience-based control into objective, data-driven predictions, significantly improving precision while the system remains accessible through standard operational interfaces
Data Source
AI summary
Provided are a device and method for predicting product properties of a naphtha splitting unit (NSU). The method includes training a prediction model for predicting the product properties of the NSU, inputting an input variable to the trained prediction model to acquire a prediction value for each output variable, and outputting the acquired prediction values for the output variables.