Distillation Column Flooding Prediction Using Synchronized Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting flooding in distillation columns are inefficient, leading to high rates of false positives and false negatives, resulting in costly production slowdowns and prolonged stabilization times.
Innovation Solution
A machine learning-based prediction method that constructs and trains a model using real-time data from sensors, incorporating data synchronization and derivative calculations to improve prediction accuracy, reducing delays and enhancing the relevance of data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional prediction methods based on theoretical equations are used, then the prediction system is simple to implement, but the prediction accuracy is low with high rates of false positives and false negatives
Solution Approach 1:
The patent replaces conventional theoretical equation-based prediction methods with machine learning models that process sensor data. The system uses trained machine learning models to predict flooding conditions, substituting traditional mechanical/mathematical approaches with data-driven intelligent systems that achieve higher accuracy while managing complexity through automated training and deployment.
2Reliability
If conventional predictors are used, then the system is easy to operate, but the number of erroneous predictions is high leading to costly production slowdowns
Solution Approach 1:
The machine learning models are trained offline using historical data and then deployed to automatically predict flooding conditions without requiring continuous manual intervention. The system self-adjusts and improves through automated retraining with new data, reducing operational complexity while enhancing reliability through continuous learning from actual plant conditions.
3Loss of time
If conventional detection methods are used, then the detection process is simple, but the detection time is delayed occurring after flooding has already happened
Solution Approach 1:
The machine learning models analyze multiple sensor parameters simultaneously to predict flooding conditions before they occur. By processing data from temperature, pressure, flow rate, and other sensors in real-time, the system provides advance warning of impending flooding, enabling preventive action before the actual flooding event happens, thus reducing detection time and avoiding production losses.
Data Source
AI summary
The invention relates to a method for predicting flooding in a distillation column by machine learning including a constructing and training phase of a machine learning model obtained from previously collected data and from a set of sensors, an operational phase for predicting flooding(s), by collecting a current data flow until a buffer is filled, pre-processing data from the data buffer by predetermined cleansing and classification, synchronizing the data of the current set of clean and classified data, determining a value of a current variable representative of at least one current performance of the at least one distillation column, forming a current set of transformed data by calculating predetermined derivatives, and predicting the current state of said distillation column by applying said learning model to said current set of transformed data.