ATM PSU Degradation Detection Using Terminal Event Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated teller machines (ATMs) face challenges in detecting power supply unit (PSU) degradation due to the lack of sensors and the slow, intermittent nature of the degradation process, leading to unnecessary part replacements and prolonged downtime.
Innovation Solution
A machine-learning model (MLM) is trained on labeled event types and patterns associated with PSU failures, using historical maintenance records to predict the likelihood of PSU issues or failures, enabling proactive detection and classification of terminals as low, medium, or high risk.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor-based detection methods are used for PSU degradation, then direct detection of power supply issues is possible, but no existing sensor can detect PSU degradation issues and the degradation process is slow and evolves over time
Solution Approach 1:
The patent uses an intermediary machine learning model that indirectly detects PSU degradation by analyzing patterns in terminal events and maintenance records, rather than directly measuring PSU parameters with sensors. This mediator approach enables detection without requiring direct PSU sensors.
Solution Approach 2:
The patent replaces the mechanical/sensor-based detection system with a data-driven machine learning system that processes terminal events and maintenance records to predict PSU failures, substituting physical sensing with computational analysis.
2Ease of repair
If field technicians replace parts based on intermittent jam symptoms, then immediate symptom relief is achieved, but wrong parts are replaced and multiple trips are needed to resolve the real issue
Solution Approach 1:
The patent performs preliminary detection and classification of PSU degradation risk before failures occur, enabling proactive maintenance scheduling. This preliminary action prevents the need for multiple reactive field trips by identifying at-risk terminals in advance.
Solution Approach 2:
The patent implements a feedback loop where terminal events and maintenance records are continuously analyzed by the machine learning model to update risk classifications, creating a learning system that improves maintenance decisions over time and reduces unnecessary trips.
3Reliability
If proactive detection of PSU degradation is implemented, then failures can be predicted and prevented, but no existing method can detect PSU degradation due to lack of sensors and slow deterioration
Solution Approach 1:
The patent creates a universal machine learning model that can detect PSU degradation across different terminal types by analyzing common event patterns and maintenance records, making the detection capability applicable to various ATM configurations without terminal-specific sensors.
Solution Approach 2:
The patent changes the detection parameters from direct PSU measurements to terminal-level event patterns and maintenance history, transforming the detection problem into one that can be solved with available data and machine learning techniques.
Data Source
AI summary
Event streams of terminals for a given interval of time are preprocessed to label event types and label predefined time-based or sequence-based patterns associated with terminal power supply unit (PSU) failures. The labeled event streams are provided as input to a trained machine-learning model (MLM), which outputs a score for each terminal representing a likelihood that the corresponding terminal is or is not going to experience a PSU failure. In an embodiment, each score is compared against one or more threshold values and each terminal is classified as low risk, medium risk, or high risk of a PSU failure. In an embodiment, the scores and/or the classifications for the terminals are reported to an enterprise associated with the terminals at predefined intervals of time.