Adding early warning system to CNN and LSTM RUL prediction algorithm

I am trying to implement an early warning system to the RUL prediction algorithm (Remaining Useful Life Estimation using Convolutional Neural Network example posted on Mathworks). I would like to make a fuzzy logic based early warning system. (ie. <0.3 = normal operation, 0.5-0.7 = medium warning, >0.7 high warning). May I know if there is any recommended method to implement this warning system?