We introduce a non-parametric estimation of the trimmed regression by using the local linear method of a censored scalar response variable, given a functional covariate. The main result of this work is the establishment of almost complete convergence for the constructed estimator. A simulation study is carried out to compare the finite sample performance based on the mean square error between the classic local linear regression estimator and the trimmed local linear regression estimator. Moreover, a real data study is used to illustrate our methodology.