A huge percentage of published scientific research can’t be dupplicated, of course the social sciences are the worst offender. For the social sciences I think I have seen figures as high as 75% unrepeatable. Biological sciences have had some huge scandals. A fair amount of fraud exists but I think the majority of it is due to sloppiness driven by academic “publish or perish”.
Sloppiness is ineptitude, perhaps, but it isn’t fraud. Academic fraud means fudging your data, or counterfeiting data, or stealing someone else’s research and publishing it as your own [but it’s not a professor putting his name on a graduate student’s research or leaving student’s name off of something where his contribution was perfunctory performance of an assigned task.