A signal processing method for reflective fiber optic displacement sensor is presented by means of a differential evolution optimized extreme learning machine (DE-ELM). The sensing head of the sensor is a combination of an illuminating fiber bundle transmitting incoming light beam and two receiving fiber bundles employed to collect the reflected beam from the reflector. Three fiber bundles with same type are put together and arranged side by side, but the two receiving fiber bundles enfaces have different distances from the reflector surface. The DE-ELM is used for extending the measuring the range of reflective fiber optic displacement sensor. A simulation experiment has been illustrated. The experimental results show that the measuring range can be extended to the whole response characteristics of the fiber optic displacement sensor and a high measuring accuracy can be obtained by the proposed method.
An intelligent optimization technique based on particle swarm optimization is proposed to identify parameters of chaotic
optical systems. The feasibility of this approach was demonstrated with the computer simulation through identifying the
parameters of Bragg acoustic-optical bistable system. The performance of the particle swarm optimization technique was
compared with the more common genetic algorithm in terms of parameter accuracy and computation time. Simulation
results demonstrated that the particle swarm optimization has better performance than the genetic algorithm in solving
the parameter identification problem of chaotic optical systems.
A new method for position error correction of position-sensitive detector (PSD) using least squares support vector machine (LS-SVM) is presented. The LS-SVM is established based on the structural risk minimization principle rather than minimize the empirical error commonly implemented in the neural networks, LS-SVM achieves higher generalization performance than the MLP and RBF neural networks in solving these machine learning problems. Another key property is that unlike MLP’ training that requires non-linear optimization with the danger of getting stuck into local minima, training LS-SVM is equivalent to solving a set of linear equations. Consequently, the solution of LS-SVM is always unique and globally optimal. A difference with the RBF neural networks is that no center parameter vectors of the Gaussians have to be specified and no number of hidden units has to be defined because of Mercer's condition. The position error correction procedure has been illustrated using 2D PSD as example. The results indicate that this approach is effective, and the position detection errors can be reduced from ±300μm to ±10μm.
A modeling and non-linear correction method for the two-dimension photoelectric position sensitive detector (PSD) is presented by means of a radial basis function (RBF) neural network. Utilizing its powerful ability in function approximation, the RBF network can perform the mapping between the PSD's readings and the light spot actual position. In order to obtain the mapping, the RBF network is trained by learning algorithm with the input/output data pairs of the PSD. The mapping is used as an inverse model of the PSD from the readings to the light spot actual position or as a forward model of it from the light spot actual position to the readings. The inverse model based on RBF network is used as a corrector. This model provides a linear response when the PSD's readings applied to the inputs of the RBF network during operation. The example shows that the measuring system with a proper RBF network correction can provide a high linearity over a wide position range. Furthermore, the forward model that expresses the characteristics of the PSD will be beneficial to provide the theoretical instruction for the analysis, design and application of the PSD.