Piston Error Measurement for Segmented Telescopes Based on a Hybrid Artificial Neural Network

Area of Science:

• Optical Engineering
• Astronomy Instrumentation
• Artificial Intelligence in Optics

Background:

• Segmented telescopes require precise alignment of sub-mirrors to avoid piston errors.
• Existing piston error detection methods are often complex and difficult to implement.
• Accurate measurement of piston errors is crucial for optimal telescope performance.

Purpose of the Study:

• To develop a new, accurate, and wide-range method for measuring piston errors in segmented telescopes.
• To leverage artificial intelligence, specifically hybrid neural networks, for improved piston error detection.
• To reduce the complexity and hardware costs associated with piston error measurement.

Main Methods:

• A hybrid artificial neural network combining Resnet and BP networks is proposed.
• Resnet learns the relationship between focal plane images and piston error indicators.
• BP network learns the relationship between modulation transfer function (MTF) and piston error magnitude.

Main Results:

• The hybrid network accurately detects piston errors using only focal plane images of a point source.
• Achieved a high detection accuracy of 10 nm.
• Demonstrated a wide detection range, covering the entire coherent length of broadband illumination.

Conclusions:

• The proposed hybrid neural network method offers a robust solution for piston error measurement.
• The method provides high accuracy, a broad detection range, and is cost-effective.
• This approach simplifies piston error detection, benefiting segmented telescope calibration and performance.