Fractional Design Generation for Multilevel Factor Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for stability testing in various industries, such as pharmaceuticals and mass spectrometry, require a large number of costly and time-consuming experimental trials due to the need to test all combinations of factors, which is not feasible with existing algorithms that primarily work with systems involving only 2 levels, leading to a demand for more efficient methods handling multilevel factors.
Innovation Solution
A computer-implemented method and system that generate orthogonal or nearly orthogonal and balanced experimental designs by decomposing factors into level sets and using Latin square matrices to create orthogonal arrays, which are then mapped to form active matrices for efficient testing of product characteristics across multiple partitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all combinations of factors are tested to ensure complete product characteristic evaluation, then measurement precision and reliability are improved, but the number of experimental runs increases significantly
Solution Approach 1:
The patent segments the experimental design into multiple partitions, where each partition tests a specific subset of factor combinations. Instead of testing all combinations in one exhaustive set, the design divides the experimental space into manageable partitions that collectively cover the full range of factor levels across multiple runs, reducing the number of trials needed at any single partition point.
Solution Approach 2:
The patent introduces the dimension of time by conducting experiments across multiple partitions (e.g., different time points or stages). By distributing the testing of factor combinations across these temporal partitions, the system achieves comprehensive evaluation without requiring all combinations to be tested simultaneously, thereby reducing the experimental burden at each partition while maintaining overall measurement precision.
2Ease of operation
If traditional algorithms are used that work with only 2 levels, then ease of operation is maintained, but adaptability to multilevel factors is limited
Solution Approach 1:
The patent changes the parameter structure by representing multilevel factors as multiple binary factors. Each multilevel factor with L levels is decomposed into multiple binary factors that together encode the L levels. This transformation allows traditional binary-level algorithms to handle multilevel factors effectively, maintaining algorithm simplicity while achieving versatility in handling complex factor structures.
3Loss of time
If the number of experimental runs is reduced to lower costs, then loss of time and resources is decreased, but measurement precision may be compromised
Solution Approach 1:
The patent performs preliminary decomposition of multilevel factors into binary representations before conducting experiments. This preliminary action structures the experimental design in advance, allowing the system to efficiently determine which factor combinations to test at each partition. By pre-processing the factor structure, the system optimizes the experimental runs to achieve maximum measurement precision with minimal trials.
Solution Approach 2:
The system uses feedback from previous partition results to inform subsequent experimental decisions. By analyzing data from earlier partitions, the algorithm can identify patterns and relationships that reduce the need for exhaustive testing in later partitions, thereby maintaining measurement precision while reducing overall experimental duration and resource requirements.
Data Source
AI summary
A method and system for creating a design plan to test a product characteristic are described. One or more factors, level corresponding to the factors, and partitions for testing the product characteristic are determined. For each partition, an active matrix is generated. The product characteristic can be tested at each partition using the levels for the factors specified by the corresponding active matrix.