Data Analytics System Optimizes KPIs via Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial processes face challenges in identifying and optimizing key performance indicators (KPIs) due to the large number of variables and complex nonlinear interactions, with existing data analytics methods failing to effectively select relevant variables and provide comprehensive optimization solutions.
Innovation Solution
A system and method that analyzes data from industrial processing units, involving data reception, fusion, verification, pre-processing, integration with physics-based models, regime identification, feature selection, predictive model building, and optimization using techniques like gradient search and clustering to optimize KPIs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data analytics methods are applied to analyze industrial data, then optimization of KPIs is achieved, but the complexity of variable selection and analysis increases significantly
Solution Approach 1:
The patent segments the complex variable selection process into multiple distinct modules including univariate analysis module, correlation analysis module, feature selection module, and predictive model building module. Each module handles a specific aspect of analysis, breaking down the overwhelming complexity into manageable, sequential steps that systematically identify relevant variables from thousands of process parameters.
Solution Approach 2:
The patent introduces an intermediary layer of data preprocessing and feature engineering that bridges raw industrial data and predictive models. This intermediary process includes data cleaning, normalization, handling missing values, and feature transformation, which simplifies the input data structure before it enters the complex optimization algorithms, thereby reducing the overall system complexity.
2Measurement precision
If comprehensive data from multiple sources is integrated, then accuracy of predictive models is improved, but data processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary data processing actions before model training, including data cleaning, normalization, handling missing values, and feature engineering. By preparing the data in advance with appropriate transformations and selections, the system reduces the computational burden during model training and deployment, thereby reducing overall processing time while maintaining model accuracy.
Solution Approach 2:
The patent extracts and focuses on the most relevant features and variables from the comprehensive industrial dataset using correlation analysis and feature selection techniques. By taking out only the most significant variables that contribute to predictive accuracy, the system reduces the dimensionality of the data, thereby decreasing computational resources and processing time required for model training and inference.
3Manufacturing precision
If advanced analytics techniques are used to identify relevant variables, then quality of optimization results is improved, but expertise and time required for implementation increase
Solution Approach 1:
The patent implements self-service capabilities where the system automatically performs variable selection, feature engineering, and model optimization without requiring extensive manual intervention or expert knowledge. The automated pipelines include built-in heuristics and algorithms that autonomously identify relevant variables and tune model parameters, making advanced analytics accessible to users without specialized expertise while maintaining high optimization quality.
Solution Approach 2:
The patent incorporates feedback mechanisms that automatically evaluate model performance and adjust variable selection criteria accordingly. The system uses cross-validation, performance metrics, and iterative refinement to automatically tune the analytics process, providing feedback loops that improve implementation ease by eliminating the need for manual expert tuning while maintaining high-quality optimization results.
Data Source
AI summary
A system and method for performing data-based optimization of performance indicators of process and manufacturing plants. The system consists of modules for collecting and merging data from industrial processing units, pre-processing the data to remove outliers and missingness. Further, the system generates customized outputs from data and identifies important variables that affect a given process performance indicator. The system also builds predictive models for key performance indicators comprising the important features and determines operating points for optimizing the key performance indicators with minimum user intervention. In particular, the system receives inputs from users on the key performance indicators to be optimized and notifies the users of outputs from various steps in the analysis that help the users to effectively manage the analysis and take appropriate operational decisions.