The answer to this question is not easy. I will provide a short outline. Some aspects of engineering are well established and well known. The Fourth Industrial Revolution is just a slogan, it has no established concepts or principles. Nevertheless, if some aspect of engineering can be codified within a sound logical framework, then is it feasible and in some cases easy to create what can be characterized as "domain specific artificial intelligence" A simple example is the task of planning the sequence of steps for assembly of mechanical device based upon analysis of precedence relations and functional dependencies. The principles that enable a person to determine a sequence of steps can be implemented in computer programs. So I would say that established engineering methods and principles are the foundation for domain specific artificial intelligence when combined within a formal framework. I would be happy to provide some examples.
To classify mechanical parts using machine learning or deep learning which could enable us to recommend the mechanical parts from a standard library based only on an image or a CAD model.
As mechanical engineering deals with robotics design that eventually connects with AI. Advances in automotive safety through Fourth Industrial Revolution technologies can reduce road fatalities and insurance costs, and carbon emissions. Digital technology can liberate workers from automatable tasks, freeing them to concentrate on addressing more complex business issues and giving them more autonomy.
As Qamar Ul Islam specified, mechanical engineering has to do with robotics, which obviously interconnects with artificial intelligence.
In the same direction, I would highlight an extremely interesting and current technical field in which mechanical engineering is also used: Microelectromechanical systems (MEMS). See the wikipedia article (https://en.wikipedia.org/wiki/Microelectromechanical_systems) for an introduction to the field.
Also I would also mention an extremely widespread artificial intelligence technology: pattern recognition. It can be used successfully for automatic identification of various mechanical components; here is an article which exemplifies it: Artificial Intelligence in Mechanical Engineering (https://towardsdatascience.com/artificial-intelligence-in-mechanical-engineering-a9dd94adc492).
Above all, see the book Artificial Intelligence in Mechanical and Industrial Engineering by Kaushik Kumar, Divya Zindani, J. Paulo Davim.
How can mechanical engineering knowledge be linked to the Fourth Industrial Revolution and Artificial Intelligence (AI) >>>> unfortunately, discussion of AI capabilities is all too often explained in an anthropomorphic fashion. At this point in time, use of the word KNOWLEDGE appeals to our imagination but does not have a REPRESENTATION suitable for implementation in what we may call AI (which itself has no agreed upon definition). These are reasons why I recommend that we focus on domain specific artificial intelligence based upon some useful facts, a means to represent those facts, and a framework for formal reasoning using those facts. My work in scheduling generally satisfies that criteria and the scheduling systems that I have worked can create schedules that are considerably better than those that a human may create. I have other work in temporal reasoning, diagnostic reasoning, and operational reasoning that generally satisfy that criteria. I am happy to discuss in more detail of others are interested.