Cold Rolling Mill Chattering Prediction Using Frequency-Domain Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting chattering in cold rolling mills are prone to delays in prediction and can lead to reduced productivity due to frequent false vibration sign detection, especially when low vibration intensity thresholds are set, which decreases rolling velocity.
Innovation Solution
A method and device that utilize a prediction model trained with multidimensional data from past rolling records to accurately predict chattering by using a neural network model, adjusting rolling conditions such as speed and oil supply based on real-time data analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a low vibration intensity threshold is set for detecting chattering, then chattering detection sensitivity is improved, but false detection increases and rolling velocity decreases
Solution Approach 1:
The system performs preliminary analysis by converting vibration signals to frequency domain data and calculating temporal changes before making detection decisions. This preprocessing step enables early identification of chattering trends without requiring immediate threshold-based reactions, allowing for more accurate detection while avoiding premature speed reductions
Solution Approach 2:
The patent introduces an intermediate processing stage that transforms raw vibration intensity data into frequency domain characteristics and temporal change rates. This intermediary representation serves as a bridge between raw sensor data and detection decisions, enabling more nuanced discrimination between normal vibration and actual chattering conditions
2Reliability
If vibration sign detection is used to prevent chattering, then chattering prevention is achieved, but prediction delay occurs due to rapid chattering occurrence
Solution Approach 1:
The system calculates temporal changes in frequency domain data as an intermediate step before final detection. This preliminary calculation of rate-of-change information enables the system to anticipate chattering development before it fully manifests in raw vibration signals, reducing prediction delay while maintaining reliable prevention
Solution Approach 2:
The patent implements a feedback mechanism where the calculated temporal changes in frequency domain characteristics are continuously monitored and fed back into the detection logic. This continuous feedback loop enables real-time adaptation and early warning of chattering conditions, allowing for timely intervention before full chattering occurs
Data Source
AI summary
A method of detecting chattering in a cold rolling mill, the method includes: a step of predicting occurrence of chattering during rolling of a material to be rolled, by inputting second multidimensional data to a prediction model, the second multidimensional data having been generated based on condition data corresponding to array data related to the material to be rolled, and the prediction model having been trained with an explanatory variable and an objective variable, the explanatory variable being first multidimensional data generated based on one-dimensional array data representing a past rolling record of rolling of rolled materials by means of a cold rolling mill, and the objective variable being a past record of occurrence of chattering corresponding to the past rolling record.