Encoder Device Using Neural Network Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Encoder devices face challenges in accurately determining kinematic values, such as position, speed, and acceleration, due to environmental and mechanical influences, requiring complex calibrations and adjustments, and existing evaluation methods are not robust against unknown or varying conditions.
Innovation Solution
The use of machine learning, specifically deep neural networks, to evaluate scanning signals from encoder devices, allowing for flexible and robust determination of kinematic values without the need for detailed system knowledge or extensive calibration, by training with simulated or measured data and accounting for various influences during training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional mathematical evaluation methods (e.g., arctangent function) are used for sin/cos code tracks, then position determination is achieved, but the system architecture becomes complex requiring acquisition, compilation, interpolation, and correction of different partial pieces of information
Solution Approach 1:
The patent combines multiple code tracks (absolute track, incremental track, sin/cos track) into a unified evaluation approach using neural networks. Instead of processing each track separately through complex mathematical operations, the neural network integrates all tracking signals simultaneously, merging the evaluation functions into a single robust system that determines position without requiring separate acquisition, compilation, and correction steps for each track type.
Solution Approach 2:
The patent replaces conventional mathematical evaluation methods (arctangent functions, interpolation algorithms) with a neural network-based evaluation system. This substitution transforms the mechanical/mathematical processing system into an intelligent system that learns optimal evaluation strategies during operation, simplifying the system architecture while maintaining or improving measurement precision through adaptive pattern recognition.
2Measurement precision
If complex calibrations and adjustments are performed to account for environmental and mechanical influences, then measurement accuracy is improved, but the process requires detailed system knowledge and extensive time
Solution Approach 1:
The patent implements preliminary action by training the neural network during a setup phase with diverse training data that encompasses various environmental conditions and mechanical influences. This preliminary training equips the system with pre-learned compensation strategies for temperature variations, mechanical tolerances, and other disturbances, eliminating the need for time-consuming on-site calibrations and adjustments while maintaining high measurement accuracy.
Solution Approach 2:
The neural network enables the encoder system to self-adjust and self-optimize its evaluation parameters during operation. Instead of requiring external experts to perform detailed calibrations, the system automatically adapts to environmental and mechanical influences through its learned evaluation models, making the calibration process self-service and eliminating the need for detailed system knowledge from operators.
3Measurement precision
If complete correction for all environmental and mechanical effects is attempted, then measurement accuracy would be improved, but complete correction is often not possible because the effects are not sufficiently known
Solution Approach 1:
The patent employs parameter changes by using neural networks that can dynamically adjust evaluation parameters based on learned patterns from training data. Instead of attempting to model and correct each individual environmental and mechanical effect separately (which would be incomplete), the neural network learns the combined effect of multiple parameters simultaneously and adapts its evaluation strategy accordingly, achieving reliable correction even when individual effects are not fully understood.
Solution Approach 2:
The neural network evaluation system provides universal correction capability that handles multiple types of environmental and mechanical influences through a single unified model. Rather than requiring separate correction mechanisms for each type of interference (temperature, vibration, mechanical tolerances), the multi-functional neural network learns to compensate for all these effects simultaneously, achieving complete and reliable correction across diverse operating conditions.
4Ease of manufacture
If a simple and cost-effective encoder device design is used, then manufacturing costs are reduced, but robustness against environmental and mechanical variations may be compromised
Solution Approach 1:
The patent replaces complex mechanical correction mechanisms (precision adjustment components, temperature compensation hardware, vibration isolation systems) with a neural network-based software solution. This substitution allows the use of simpler, more cost-effective hardware designs while maintaining or improving robustness against environmental and mechanical variations through intelligent evaluation and adaptive compensation performed by the neural network.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables a simple and cost-effective encoder device design with improved accuracy and robustness against environmental and mechanical variations, reducing measurement errors and eliminating the need for complex calibration, while allowing for precise determination of kinematic values.
Implementation Method 1
Optical rotary encoders use a code disk having openings or reflection marks that serve as the code and that modulate the signal of a light transmitter so that a light receiver arranged in transmission or reflection receives position signals
Implementation Method 2
With a magnetic encoder, the change of the magnetic field on the moving past of encoding permanent magnets is, for example, detected using a Hall sensor
Implementation Method 3
Further known systems use standards and scans that work, for example, inductively
Implementation Method 4
Further known systems use standards and scans that work, for example, inductively, capacitively
Data Source
AI summary
An encoder device for determining a kinematic value of the movement of a first object relative to a second object is provided, wherein the encoder device comprises a standard associated with the first object and at least one scanning unit associated with the second object for producing at least one scanning signal by detection of the standard and a control and evaluation unit that is configured to determine the kinematic value from the scanning signal. The control and evaluation unit is here further configured to determine the kinematic value by an evaluation of the scanning signal using a method of machine learning, with the evaluation being trained with a plurality of scanning signals and associated kinematic values.