Feature Scaling Both train and test data reduces accuracy while scaling only one of the data increase accuracy drastically

Question

If I apply feature scaling on both test and train data using the same standards scaling the accuracy is around 83% but when I scale only train data then the accuracy is arou