In laser-driven ion acceleration experiments, the time-of-flight (TOF) technique based on diamond detector serves as a key diagnostic approach for measuring the energy spectrum of accelerated ions. However, the transient electromagnetic pulse (EMP) generated during the interaction between intense laser pulse and solid target can strongly interfere with the data acquisition system, leading to significant baseline distortion in the oscilloscope signals. Such distortions may contaminate or even obscure the ion signals, posing serious challenges to accurate spectrum measurement.

This study aims to characterize the EMP induced baseline distortion in diamond detector TOF measurements and develop an adaptive correction algorithm to recover baseline to accurate ion energy spectra from contaminated single-shot data.

We developed a machine learning assisted time varying polynomial baseline correction method. The algorithm employs a segmented fitting strategy. Additionally, an adaptive moving window selection for dynamic optimization of reference point identification is introduced, with the window width adjustable from 20 ns to 10 ns.

The results show that intense EMP generated at the moment of laser-target interaction couple into the diagnostic system through the transmission cables, inducing baseline drops up to −5 V, which gradually recover to the normal level after approximately 200 ns. Polynomial orders are assigned region-specifically: first-order for instantaneous interference I and II region, third-order for continuous interference region, and sixth-order for stable recovery region. Model accuracy is validated through root mean square error (RMSE). After correction, previously obscured TOF peaks for protons and carbon ions (C¹⁺ to C⁶⁺) became clearly identifiable, enhancing the detection of low-energy ions.

This study presents an adaptive baseline correction method, which effectively reduces the EMP interference on the baseline in laser-driven ion acceleration diagnostics. The proposed model reasonably characterizes the temporal evolution of the baseline and provides a feasible approach for future online interference correction of single-shot ion TOF spectra.