Neural State Prediction for Cyclic Production Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems are unable to predict the state of system components in cyclically producing installations, such as presses for motor vehicle parts, which can lead to unscheduled stoppages and resource inefficiencies.
Innovation Solution
An apparatus and method utilizing an artificial neural network to capture and analyze time series sensor data, allowing for state predictions and anomaly detection, thereby preventing partial or complete system failures by identifying abnormal behavior in system components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional control optimization methods are used, then control variables can be optimized, but state prediction of system components cannot be achieved
Solution Approach 1:
The patent replaces traditional mechanical control optimization systems with an artificial neural network-based prediction system. The neural network processes sensor data to predict system component states, substituting conventional control methods with intelligent algorithms that can forecast failures and anomalies before they occur.
Solution Approach 2:
The patent introduces sensor data as an intermediary between the physical system components and the prediction algorithm. Sensors capture operational parameters (pressure, temperature, position) which serve as input features for the neural network, enabling indirect observation and prediction of system component states without direct intervention.
2Measurement precision
If continuous monitoring of all system components is implemented, then state prediction accuracy improves, but resource consumption increases
Solution Approach 1:
The patent applies partial monitoring by selecting only the most relevant sensor data features for prediction. Instead of continuously monitoring all possible system parameters, the neural network processes a curated subset of critical features (pressure, temperature, position) that provide sufficient prediction accuracy with reduced computational and energy resources.
Solution Approach 2:
The patent transforms raw sensor data into meaningful prediction targets by changing parameters from continuous monitoring to discrete state predictions. The system predicts specific failure states or anomaly conditions rather than continuously tracking all parameters, reducing computational burden while maintaining prediction effectiveness.
3Reliability
If comprehensive sensor data collection is performed, then prediction reliability improves, but data storage requirements increase
Solution Approach 1:
The patent extracts only the essential features from comprehensive sensor data for prediction purposes. The neural network processes selected features (pressure, temperature, position) rather than storing and analyzing complete raw sensor datasets, separating critical predictive information from redundant data to reduce storage requirements while maintaining prediction reliability.
Data Source
AI summary
An exemplary aspect of the present invention relates to an apparatus for predicting the state of at least one system component of a cyclically producing installation, comprising a recording unit for recording a time series of sensor data relating to the at least one system component of the cyclically producing installation and/or event data relating to the at least one system component, a calculation unit having at least one artificial neural network that is implemented therein and is intended to calculate a state prediction for the at least one system component on the basis of the time series of sensor data and/or event data, and an output unit for outputting the calculated state prediction for the at least one system component.