The invention disclose an auxiliary determination method for non-destructive detection of a tunnel lining in combination with machine learning, comprising steps of extracting characteristic parametersfrom the collected shock vibration signals; performing a regression fitting on a reflection time in the characteristic parameters to obtain a calibration value; marking the set of characteristic values with the obtained characteristic parameters represented as a original signal, recording the defect condition as a training set; repeating the above steps on different test objects to increase the number of training sets; training the model using a model training software; parsing the data having unknown detection results by the established model. The auxiliary determination method of the invention eliminates the adverse effects caused by change of thickness and material, increases the reflection time of the back of the lining as the defect determination parameter, so as to preferably reflect the defect characteristics, and solve the problem such as significant influence of subjective factor from a staff on the existing detection methods and less detection accuracy, achieve the effects of improving detection accuracy, reducing subjective interference from a staff, and ensuring objective and accurate detection results.