Printing Press Anomaly Detection Using Multivariable Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods in printing presses and print-processing machines are ineffective due to varying production states and materials, leading to incorrect conclusions and increased downtimes, as they only consider individual variables without accounting for relationships between them.
Innovation Solution
A computer-implemented method that incorporates multiple variables for anomaly detection, analyzing the first variable in context with at least one second variable to define normal profiles and detect anomalies, using artificial intelligence to filter out irrelevant fluctuations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If univariate anomaly detection is used to examine only one variable in isolation, then the evaluation speed is high and computing capacity requirements are low, but the detection accuracy is poor because relationships between different variables and machine states cannot be taken into account
Solution Approach 1:
The patent segments the anomaly detection process into two distinct phases: (1) an offline training phase where a multivariate model learns normal relationships between variables during normal operation, and (2) an online detection phase where only residual anomalies are detected. This segmentation allows the system to use complex multivariate analysis for learning while maintaining simple real-time operation, resolving the contradiction between detection accuracy and system complexity.
Solution Approach 2:
The system performs preliminary action by collecting and analyzing data during normal machine operation to establish a baseline model of variable relationships before any anomalies occur. This pre-learning phase enables the system to understand normal correlations between variables, so that during actual operation, only deviations from this established baseline need to be detected, improving accuracy without requiring continuous complex multivariate computation.
2Measurement precision
If multivariate anomaly detection is used to examine two or more variables simultaneously for anomalies, then the detection accuracy improves by considering inter-variable relationships, but the evaluation speed decreases and computing capacity requirements increase
Solution Approach 1:
The patent extracts and separates the complex multivariate analysis from real-time operation by moving it to an offline training phase. During online operation, only the simple residual calculation is performed, extracting the computationally intensive learning process from the time-critical detection process. This resolves the contradiction by applying multivariate analysis only when computing resources are abundant (offline) rather than when speed is critical (online).
3Measurement precision
If anomaly detection is performed without considering different production states and consumables, then the system operation is simple, but the detection accuracy decreases because individual variables can vary greatly due to different production requirements and materials
Solution Approach 1:
The patent creates a universal anomaly detection system that handles multiple production states, substrates, and consumables through a single multivariate model. The model learns the specific relationships between variables for each production condition during the offline training phase, making the system adaptable to different printing materials and machine states without requiring separate detection algorithms for each case, thus achieving universality while maintaining accuracy.
Solution Approach 2:
The system accommodates different production states by allowing the multivariate model to learn parameter-specific relationships during training. When production conditions change (different substrates, speeds, or consumables), the model captures these parameter variations in the baseline relationships, enabling accurate anomaly detection across diverse operating conditions without manual reconfiguration.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Computer-implemented method for anomaly detection. The invention relates to a computer-implemented method for evaluating data, the method comprising: receiving a data set (20) from at least one component of a printing press or a print-processing machine, the data set (20) comprising a first variable (1) with a plurality of first data points (11) and at least one second variable (2) with a plurality of second data points (12); performing computer-implemented anomaly detection of the first data points (11) of the first variable (1) to determine at least one anomaly. The invention is therefore based on the object of finding a solution in which the anomaly detection can be applied to different production states and when using different consumables.The object is achieved according to the invention in that at least the second data points (12) of the second variable (2) are taken into account in the computer-implemented anomaly detection of the first data points (11) of the first variable (1).