1.Gather the following information: what you already know; sources of relevant data; your assumptions; what you'd like to predict with the model; ways of verifying that the model will be built correctly; and ways to validate the model. Simply, read the problem many times, classify knowns and unknowns and find out what is actually asked in problem.
2.Make a strategy. After classifying the data, make the strategy how to solve the problem or how to make model. Sketch simple diagrams that outline the elements in the model and how they are connected to each other. As for any complex task, diagram helps.
3.Conduct a thorough literature review. There is no need to re-invent the wheel if somebody else has developed a model that may suit your purposes already. However, you need to fully understand all the assumptions and the applicability of a model before using it.
4.Learn Data Handling.It is important to know what is missing information in the problem. So think carefully about how you are going to handle missing data. If possible, quantify the uncertainties associated with the data. Sometimes, we overlook the missing information,so gain read problem several times and carefully.
5.Begin with a simple model. Make possibilities of different applicable models and then choose the best and simple.According to Occam's Razor principle, among models with similar predictive power, the simplest one is the most desirable.
6.Identify the parameters of the equations and develop a plan how to estimate the parameters from the data. This could be done simply by fitting the equations to the data.
7.Validate your model against a data set that you have not used to build the model.
8.Constantly test your model and update your equations based on new data and information.
9. Use Matlab and Simulink - Simulation and Model-Based Design.