Chemical engineering educators and corporations are constantly trying to decide how to respond to the evolution of computing capabilities. The educators continually review and survey to determine how much and which computing subjects to include in their curricula. Companies have to decide how long to maintain in-house computing capabilities vs. commercial software and out-sourced consulting. You may see results of those surveys and studies at cache.org.
One of the problems in the past was that many of the professors received their education prior to the maturation of the computing capabilities of the hardware and software. That is no longer the case. Now the problems relate to the transitions from one state of computing capability to another. These states are characterized by the availability of commercial programs for new applications. Example applications are process simulation, advanced process control and optimization, computational fluid dynamics (CFD), and general purpose mathematical programs (e.g. Mathcad, Matlab, Mathematica).
Transition problem 1: Should numerical methods be required study?
Commercial software now uses good numerical methods for solving ordinary and partial differential equations and models. Is a course in numerical methods still needed by ChE's? My answer is "no", but not for the reason given above. I still found a need to write my own numerical integration routine for a 2D reactor model using Mathcad even though it has a PDE solver. My answer is "no" because engineers that still need to use numerical methods can learn, as I did, by self-study. In my case, I recall being taught only simple Runge Kutta methods for ODEs in my course work. Later, I studied independently using several books, including those listed at the end of this blog. As the faculty struggles with decisions about the curriculum, they need to remember that their main product is graduates that can think and reason and study on their own.
Transition problem 2: Incomplete functionality
During the transition between states, the commercial software may not solve all of the types of problems needed. For example, most commercial process simulation software can model simple ideal reactors, but they can not accurately model a reactor tube in a furnace or in a cooling shell. Thus, firms designing chemical processes will need someone with knowledge of numerical methods and the reactor balances in order to design a tube reactor.
Transition problem 3: "Shakedown" period
When a new software technology first appears, there is a good probability that there are errors in the implementation. The users need to be cautious at first. Simple problems should be devised to test different features. Old School engineers are suited for the testing because they know some of the potential problem areas. When CFD programs began to be marketed in the early '90's, I attended a sales meeting for one of the leaders in the field. During their presentation, they showed the conservation equations being solved. I noticed an error in their energy balance. I couldn't convince them on the spot, but my written correspondence later did convince them to correct the problem. My employer still purchased their program, mainly because they were the one CFD company in the field that published extensive test results using actual data gathered by independent researchers, usually at universities. Several competitors relied more on pretty colored plots of flow fields for sales presentations. Sadly, that often works.
Eventually, the new technology becomes widespread and reliable. At that point, we can safely drop some of the required courses. After all, who knows how to find the square root of a number except by pushing a button on a calculator or writing sqroot() or something similar?
Some references for numerical methods