Microbiological Risk Modeling for Aseptic Food Batch Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quality control methods for aseptic packaging struggle to accurately model microbiological data from high-yield filling machines, leading to inefficient and costly sampling and inspection processes, with a need for improved risk evaluation and reduced production disruption.
Innovation Solution
A method and system that determine a microbiological risk level by fitting a zero-inflated binomial distribution to microbiological sampling data, using cumulative relative frequencies and zero-inflation parameters to minimize square error, allowing for real-time deviation detection and improved production inspection tools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sampling and inspection methods are used for aseptic packaging quality control, then production disruption and costs are high, but measurement precision and reliability of quality assessment are insufficient
Solution Approach 1:
The patent replaces physical sampling and inspection mechanisms with a statistical modeling system. Instead of retrieving and analyzing actual microbiological samples from packages, the system uses mathematical models (zero-inflated binomial distribution) to predict quality outcomes based on process parameters, thereby eliminating production disruption while maintaining assessment accuracy
Solution Approach 2:
The patent introduces statistical distributions and risk models as intermediary tools between the filling process and quality assessment. These mathematical intermediaries allow indirect evaluation of microbiological quality without direct physical sampling, reducing production impact while providing reliable quality metrics
2Manufacturing precision
If traditional binomial distribution modeling is used for aseptic data, then device complexity is low, but manufacturing precision and accuracy of quality modeling are insufficient
Solution Approach 1:
The patent transitions from simple binomial distribution parameters to a more sophisticated zero-inflated binomial model with additional parameters (zero-inflation parameter π and probability parameter p). This parameter expansion enables the model to capture the excess zero occurrences in aseptic data, significantly improving modeling accuracy despite increased complexity
Solution Approach 2:
The patent segments the quality modeling process into distinct computational steps: calculating cumulative relative frequencies, determining zero-inflation parameters, fitting the ZIB distribution, and evaluating risk levels. This segmentation makes the complex modeling process more manageable and implementable
3Reliability
If frequent sampling is conducted to improve quality assessment reliability, then measurement precision improves, but productivity and production efficiency decrease
Solution Approach 1:
The patent performs preliminary statistical modeling and risk assessment using historical data and process parameters before actual production batches are completed. This preliminary action allows quality control decisions to be made in advance, eliminating the need for frequent interruptive sampling during production and maintaining both reliability and productivity
Data Source
AI summary
A method of determining a microbiological risk level in food, comprising determining zero-inflated binomial (ZIB) distribution parameters (π, p); i) determining cumulative relative frequencies (fo, fi, f2, . .. , fx) for a number of occurrences (0, 1, 2, x) of defective samples; ii) calculating a vector of a sub-set of zero-inflation parameters (π) of k+1 elements according to; o/o=[0, 1* fo/k, 2*fo/ k, k*fo/k]; iii) calculating a vector of a sub-set of first parameters (β) based on the sub-set of zero-inflation parameter (π); iv) for the vector pairs (p, π) in the sub-set of first parameters and the sub¬set of zero-inflation parameters, determining a square error between said cumulative relative frequencies and cumulative theoretical probabilities Px of having ≤x occurrences over N samples for a ZIB distribution; v) determining the zero-inflation parameter lt) and the first parameter (p) as the vector pair providing the least square error.