Alarm Correlation Analysis Using Market Basket Algorithms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Alarm systems in technical installations and processes generate a high volume of alarms, leading to operator confusion and potential missed critical alerts, as existing methods for optimizing alarm configurations are limited by manual processes and lack of effective use of historical data.
Innovation Solution
A method and device using data processing to analyze historical alarm messages and operator interventions through shopping basket analysis, identifying correlations and quality indicators to reduce unnecessary alarms by treating data in defined intervals and applying algorithms like APRIORI and ECLAT, and visualizing results to inform alarm suppression rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If alarm systems generate comprehensive alarm messages for all hazardous situations, then operator awareness of plant states is improved, but operator confusion increases and critical alarms may be missed in alarm deluge
Solution Approach 1:
The patent extracts and removes redundant alarm messages from the alarm system output by analyzing historical data to identify correlations. Alarms that frequently occur together or are causally related are filtered, keeping only the most critical or independent alarms to prevent operator confusion while maintaining awareness of plant states.
Solution Approach 2:
The patent combines multiple correlated alarm messages into single representative alarms or alarm patterns. By merging redundant information and presenting consolidated alarm states, the system reduces the number of individual alarm notifications while preserving comprehensive plant state monitoring capability.
2Ease of manufacture
If manual configuration methods are used for alarm systems based on process knowledge and simple statistics, then implementation simplicity is maintained, but the ability to reduce alarms using historical data and operator experience is limited
Solution Approach 1:
The patent replaces manual configuration methods with automated data processing systems that use shopping basket analysis algorithms. These computational methods automatically analyze historical alarm data to identify correlations and generate optimization recommendations, substituting human manual analysis with automated statistical processing while maintaining ease of implementation through standardized software tools.
3Ease of operation
If the number of alarms is reduced to avoid alarm deluge, then operator focus on critical alerts is improved, but essential alarms may be suppressed
Solution Approach 1:
The patent implements feedback mechanisms where alarm reduction decisions are continuously validated against historical data and operator interventions. The system monitors the effectiveness of alarm suppression rules and adjusts configurations to ensure essential alarms remain visible, using feedback loops to maintain both operator focus and reliable detection of critical conditions.
Solution Approach 2:
The patent performs preliminary analysis of alarm correlations using historical data before implementing suppression rules. By pre-identifying which alarms are redundant versus essential through shopping basket analysis, the system establishes suppression strategies in advance that are designed to maintain critical alarm visibility while reducing overall alarm volume.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method and a device suitable for carrying out the method for identifying correlations between alarm messages of an alarm system of a technical plant or a technical process and/or correlations between such alarm messages and operator interventions, using a data processing device that has access to recorded historical data relating to alarm messages and operator interventions.In this process, historical data at defined intervals are treated and analyzed as a basket of goods using the data processing device and methods of market basket analysis, and/or, using the data processing device that has access to recorded historical data relating to alarm messages and operator interventions, quality indicators are formed from correlations between alarm messages and operator interventions based on the ratio of the respective operator interventions before and after the occurrence of the respective alarm message.