Deformable Mirror Shape Control via Neural Network Radial Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for controlling deformable mirror surface shapes are complex, require extensive mathematical analysis, and are limited in generating complex surface shapes with large surface slopes or free-form surfaces, making them unsuitable for advanced applications in adaptive optics and astronomy.
Innovation Solution
A method using radial primary functions to characterize and control deformable mirror surface shapes through a neural network, simplifying calculations and enabling the generation of complex surface shapes by processing structural characteristics and control parameters, and training the neural network with sample data to output actuator voltages for precise shape control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional influence function matrix methods are used to control deformable mirror surface shape, then the control can achieve basic surface shapes, but the calculation process becomes complex and requires extensive mathematical analysis
Solution Approach 1:
The patent replaces the traditional mechanical/mathematical influence function matrix calculation system with a neural network computational system. The neural network learns the mapping between actuator voltages and surface shapes through training data, substituting complex analytical mathematics with data-driven computation that simplifies the control process while maintaining precision.
Solution Approach 2:
The patent performs preliminary action by pre-training the neural network with extensive surface shape data before actual control operations. This pre-computation phase creates a ready-to-use control model that eliminates the need for complex real-time mathematical analysis during operation, allowing rapid generation of control voltages for desired surface shapes.
2Adaptability or versatility
If existing neural network control methods are used with full surface shape data as input, then complex surface shapes can be controlled, but the calculation amount becomes large and computation is time-consuming
Solution Approach 1:
The patent extracts only the essential features from full surface shape data to create a simplified input representation for the neural network. Instead of using complete surface shape matrices as input, the method identifies and uses key characteristic parameters that capture the essential surface shape information, significantly reducing input data dimensionality and calculation amount while maintaining the ability to control complex surface shapes.
3Adaptability or versatility
If traditional control methods are used, then simple surface shapes can be generated, but complex surface shapes with large surface slopes or free-form surfaces cannot be achieved
Solution Approach 1:
The patent performs preliminary action by pre-training the neural network with a comprehensive dataset that includes diverse surface shapes ranging from simple to complex configurations. This training phase enables the network to learn the full capability space of the deformable mirror, including large surface slopes and free-form surfaces, before actual control operations begin.
Solution Approach 2:
The patent replaces traditional mathematical control methods that are limited to simple surface shapes with a neural network-based system that can handle complex surface shapes. The neural network's ability to learn non-linear relationships from training data enables precise control of complex surface configurations that traditional analytical methods cannot achieve.
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 allows for the easy implementation and precise control of complex surface shapes, including those with large surface slopes and free-form surfaces, reducing calculation complexity and enhancing the universality of deformable mirror applications.
Implementation Method 1
most deformable mirrors use piezoelectric ceramics as actuators to ensure that deformable mirrors can produce high-precision surface shape quickly and flexibly
Data Source
AI summary
A method for controlling a deformable mirror surface shape based on a radial primary function processes samples of a neural network training set, obtains elements of the neural network training set, characterizes the complex surface shape of the deformable mirror in each sample by using a radial primary function, takes characterization parameters of the surface shape in the each sample as an input of a neural network, and trains the neural network by taking a voltage of each piezoelectric ceramic corresponding to the complex surface shape as a corresponding output of the neural network. Training times of the neural network are consistent with a number of the samples. Finally, the trained neural network is obtained to verify the training effect of the neural network. According to the characterization parameters of the required surface, the neural network is used to control the deformable mirror to generate the required surface.