How should someone choose a PhD topic so that she doesn't fail?

Make sure your advisor has a good track record of graduating students in time.

Anyone just entering or outside the field won't be able to assess PhD topics with good judgement, so it's unfair when advisors fail their students by giving them bad projects. Your best bet at avoiding this is finding an advisor who is unlikely to fail students in this way.

If you find yourself in this position, a good bet is to reach out to other professors and tell them what's going on. It can feel shameful, but I've seen many success stories of people getting a new project and spinning it into enough for a PhD when things aren't working out with their initial advisor.

A PhD is awarded following submission of a thesis. It is extremely rare for a student who submits a thesis to fail. It is quite common for a student to never submit a thesis.

If the goal is simply to pass, then the key questions should be:

• What are the expectations for a thesis in my discipline? Expectations vary, but usually originality is expected.
• Will this thesis topic allow me to meet those expectations?

The one situation where a choice of topic would be likely to directly cause failure would be if the topic is blatantly not original. For example, it is found in well-known textbooks. It is much more common for a student to stop working on their thesis because they do not like the topic.

Financial and health factors are common causes of PhD non-completion.

Speaking as someone currently in the trenches, I’d advise the following general strategies for a doctoral student to maximize their chance for completion. At the very least, all these points should be considered. Also, as others have said, you won't fail a dissertation for having a hypothesis that yields a negative result – a dissertation is very much about the process not the scientific result per se.

• Develop your dissertation to play largely to your strengths, not address your weaknesses. For example, if you’re really strong at biological research but have only just learnt to code, it might not be a good idea to have a dissertation that is centered on building a software platform – even if it does target biological research as its domain.
• Choose a topic for which you’ll have expert guidance. That means your advisor and members of your committee can understand the concepts, methodology, and novelty of your work. Their advice will also be that much more helpful; they’ll be better equipped to help you navigate the roadblocks that’ll inevitably crop up.
• Do the background to make sure you’re addressing a real gap in the current literature. It pays to be a bit future thinking and aspirational; as a PhD student, one of the advantages you have is a multi-year timeframe where you can largely focus on one thing. Don't be afraid to think big and then narrow down your focus – doing so can help give you a larger sense of purpose; it can help you remember how the little thing you're working on in the moment factor into your larger vision.
• Discretize and make independent the goals of your project. This can be tough to do, but it’s a highly effective strategy and can ensure some degree of impact among your final research products. If each goal builds on the other, it's easy for your entire project to hit a roadblock after you've already invested a lot of time into it, which amplifies the risk that your entire project could fail.
• Be wary of situations and research designs that will precipitate bureaucratic delays. IRBs, data access committees, awaiting approval from distant stakeholders (timezone delays can add up!), and long duration data generation are examples of this. If at all possible, design your project to at least have a primary endpoint that won’t require more than one of these. Note that not all of these potential roadblocks are created equal. In my experience, the order of the above delays looks something like this: Long duration data generation > IRB > data access committees > distant stakeholders.
• Document communications and decisions with your committee and administration in writing. For example, when seeking input on a larger project decision from your committee members via email, be sure to state (in a friendly way) when you need a response by and the default action that will occur if no response is received by that date. Send a friendly reminder 48 hours before the date if you haven't received a response. For big decisions and reviews, allow your committee 2 weeks of lead time.
• Have an insurance policy. This is something I often setup before making a big career decision – ultimately, failure is always a possibility. What I mean by this is to have something to fallback on if your primary focus (i.e. your doctorate) ends in failure. As an example, I completed an MS prior to pursuing a doctorate and have a software side project and associated business plan that I believe are together legitimately valuable and actionable – at the very least, both would help me land a job that I would enjoy and keep me stable. Having 'insurance' can help give you peace of mind and sustained focus when pursuing something that might be inherently risky, and in some respects doctoral degrees are.

This isn't an exhaustive list; there are other considerations as discussed in other answers. That said, in my personal experience (and observing others at my institution) I’d recommend being mindful and discussing all of these aspects of your dissertation, possibly throughout your doctoral research, with your advisor and/or committee, though the latter may be best discussed with your peers.