Scientific fraud is usually in things that aren't normally checked by peer review. For instance, suppose someone writes a fraudulent paper showing that broccoli leads to homosexuality.

The peer reviewers will be looking at the design of his experiment--could it actually show such a link, and does it control for other factors that might explain the result? Does the paper give enough detail for others to repeat the experiment? Is the analysis of the data sound statistically? Is the work original? Is it citing appropriate prior researchers?

Even in a fraudulent paper, those aspects of the paper should be OK. The fraud will generally be in lying about the results of running the experiment, which is beyond the scope of peer review.

Where that gets caught is after the paper is published, and other researchers try to replicate it and get different results.