The trick isn't implementing the feature: that's basic timekeeping, file use, subtraction, etc. The trick is ensuring they don't bypass it. The body of tricks you must google for that is called "code obfuscation techniques." If this was easy, there wouldn't be entire industries producing and countering them. The major tech in this area is DRM, which is basically what you're creating. It typically doesn't work if it's on a machine they control. So, the software needs to be on one you control.
The TPM's are designed to accomplish this, but are failures. The web option another mentioned is possible, but you say that's unacceptable. The last solution that can work in standalone is to provide a piece of hardware that runs your software while protecting it. Technology such as CODESEAL[1], SecureCore, SecureME, etc (Google them) allows software to run in a system while not trusting memory, devices, etc. You might combine a tech like CODESEAL, a simple OS, a driver, and a tiny form-factor PC to make a solution somewhat convenient for your users. They plug the device into their PC, drivers for communication are installed, an app for interface is installed, they register their device via a code (or online), and from that point on they just click an icon to see the application. Even though it seems like it's on their system, it's actually running on the device which protects its code via memory encryption & control flow protection.
That was my design a while back for solving this problem.
[1] http://www.microsemi.com/products/information-assurance/softwareanti-tamper/codeseal