Cutter-Rock Interaction Modeling for Sparse Force Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current rock cutting simulators face challenges in accurately estimating cutter forces due to insufficient data on cutter/rock interactions, leading to inaccurate and inefficient rock cutting operations, as they must interpolate or extrapolate data for various cutter sizes and rock types, resulting in slower and more expensive processes with increased wear on equipment.
Innovation Solution
A computer-implemented method using neural networks and analytical models to expand limited rock cutting test data, generating synthetic datasets for different cutter sizes and rock types, improving simulation accuracy and efficiency by training models on calibrated data and applying machine learning techniques to predict cutter/rock interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional interpolation or extrapolation methods are used for cutter force calculations, then the simulator can provide estimates for various cutter sizes and rock types, but the accuracy of the estimates deteriorates due to insufficient test data
Solution Approach 1:
The patent creates virtual copies of test data through synthetic data generation. Neural networks are trained to generate synthetic cutter/rock interaction data that replicates the characteristics of actual test data, allowing the simulator to provide accurate estimates for cutter sizes and rock types beyond the limited physical test dataset without requiring extensive physical testing of every combination
Solution Approach 2:
The patent transforms the limited physical test data into expanded datasets by applying parameter variations through neural network generation. The system changes parameters such as cutter size, rock type, and operating conditions to generate synthetic data points that maintain physical consistency while expanding the coverage of the dataset, thereby improving both adaptability and accuracy simultaneously
2Measurement precision
If extensive cutter/rock testing is conducted for all cutter sizes and rock types, then the accuracy of cutter force calculations improves, but the time and cost of the process increases
Solution Approach 1:
The patent performs preliminary data expansion before actual simulation operations. By pre-generating synthetic test data using neural networks trained on limited physical test data, the system prepares a comprehensive dataset in advance that covers various cutter sizes and rock types, eliminating the need for time-consuming physical testing during operational phases
Solution Approach 2:
The patent creates virtual test data that replicates physical test results. Instead of conducting extensive physical tests, the system uses neural networks to generate synthetic copies of test data that capture the essential relationships between cutter parameters, rock properties, and cutting forces, significantly reducing testing time while maintaining accuracy
3Quantity of substance
If extensive cutter/rock testing is conducted for all cutter sizes and rock types, then the completeness of the dataset improves, but the cost of the process increases
Solution Approach 1:
The patent generates synthetic test data that complements physical test data, creating a more complete dataset without proportional increases in testing costs. The neural network-based synthesis creates virtual test results for cutter/rock combinations that would be expensive or time-consuming to test physically, thereby expanding dataset completeness while controlling costs
Solution Approach 2:
The patent systematically varies parameters in the synthetic data generation process to ensure comprehensive coverage of the parameter space. By controlling parameter variations in the neural network synthesis, the system achieves complete dataset coverage for multiple cutter sizes and rock types at a fraction of the cost of physical testing
4Reliability
If more test data is collected for different cutter sizes and rock types, then the reliability of the simulator improves, but the complexity of data collection and processing increases
Solution Approach 1:
The patent uses neural networks to generate synthetic test data that maintains the statistical and physical properties of actual test data. This copying approach increases the volume and diversity of available test data, thereby improving simulator reliability, while avoiding the operational complexities associated with conducting and managing extensive physical testing programs
Data Source
AI summary
A computer-implemented method may include receiving test data representing a cutter/rock interaction for a cutter/rock pair; calibrating an analytical model to represent the cutter/rock interaction mechanism for a cutter/rock pair; applying the calibrated analytical model to expand the test data to form one of a plurality of expanded test datasets; generating a first neural network model, of a plurality of first neural network models, representing cutter/rock interaction between a plurality of cutters of different cutter sizes and a particular rock type, wherein the first neural network is generated using the plurality of expanded test datasets as training input; generating a second neural network model using the plurality of first neural network models as training input, wherein the second neural network model represents non-tested cutter/rock interactions between a plurality of cutters of different cutter sizes and a plurality of rock types.