The key thing I learned doing sensor projects is you need to define the requirements, operational environment, acceptable failure rate, range and things like the cost of downtime, criticality and accuracy etc. With sensor based projects it is even more critical to define the requirement simply because you can find cheap crap from China and it will sorta work under ideal conditions but may fail quickly when put into a factory or warehouse. So if you are running a business off them I'd rather pay the 10x cost for a quality sensor and not be replacing it every week and know that when shit goes down it will work, especially if safety is a concern.
For playing around, element14 (newark) and adafruit have all kinds of neat sensors to play with, then I usually turn to mouser or digikey or direct from manufacturer if you need quantity. Companies like GE, Siemens, SenseAir and a huge list of others make all kinds of sensors depending on the need.
Just thinking, predictive maintenance, might mean load sensors, wear indicators, level indicators, vibration, alignment (laser) etc, so I'd start with a few ideas of what you need and then ask as it might get you better recommendations (also search element14, adafruit for ideas). Especially if you can put some tolerances to the needs as tolerances will narrow down the right vendors much quicker.