Charge Reconstruction Of Silicon Microstrip Detectors Based On Machine Learning Techniques

Silicon micro-strip detectors have the advantages of high position resolution,high energy resolution and wide linear range,and are widely used in particle detection.The signal amplitude of each readout channel of the silicon micro-strip detector is related to the incident position,incidence direction and charge quantity of the charged particles.To accurately identify the incident charge of the charged particles on a perevent basis,the signal amplitudes of each readout channel need to be utilized for charge reconstruction.However,the charge reconstruction of classical methods has problems such as inability to identify abnormal reconstruction events,high statistical requirements for scale,and insufficient use of multi-channel information,in view of the above problems,this paper introduces machine learning algorithms and proposes to optimize the classical method and PCA(Principal Component Analysis)method to improve the charge resolution of silicon micro-strip detectors.The optimization classical method makes two optimizations on the basis of the classical method: first,in view of the problem that the classical method cannot identify abnormal events,the optimization classical method proposes to denoise the data with the OPTICS(Ordering Points To Identify the Clustering Structure)clustering algorithm before the charge reconstruction to eliminate the abnormal events that affect the charge reconstruction effect;secondly,in order to reduce the manual intervention of the step of artificial charge assignment in the classical method,the optimization classical method proposes to use the Gaussian Mixture clustering algorithm to autonomously assign labels to the data.The experimental results show that based on the above two optimizations,the charge resolution of the optimized classical method is significantly better than that of the classical method,and its value is reduced by 16.5%when the charge amount Z=6.Aiming at the problem that the charge reconstruction of the classical method and the optimized classical method do not make full use of the multi-channel information,the PCA method proposes to use the PCA algorithm to analyze the variance contribution rate of each channel on the basis of the optimization classical method,and assign different weights to each channel with different signal-to-noise ratio.The experimental results show that compared with the optimized classical method,the charge resolution of the PCA method is smaller,the peak-to-valley ratio of the reconstructed charge spectrum is larger,and the charge reconstruction effect is better,and the charge resolution is increased by 10.7% when the charge amount Z=16.