The evolving breed of Mechanical Engineers?!

Our batch of 1998 consisted of 110 mechanical engineers (including one lady engineer, I must add). 14 years since, today, there are hardly 10 to 15 of these engineers in the field that they were trained for. Of course, those select few have moved over to managerial and advisory roles, but are still in the domains of mechanical engineering. The majority rest were diluted by the lure of opportunities offered by the information technology and management world. For example, the starting salary offered to a fresh ME in 1998 was about Rs. Ten Thousand a month by mechanical engineering companies as compared to about Rs. 20-25 thousand a month for an ME who had got into the software stream. One brave soul asked the employers of a large software consultancy firm at their PPT (Pre-Placement Talk) the reason they were hiring engineers from other branches. They promptly said that they were looking at the maturity and intelligence absorbed during the four years of engineering, all the training required to prepare the recruits for the industry would be provided by them in 3 months, based on the foundations of engineering in any branch!

Any of the elite IITs would spend around Rs. 200,000 per year to train each of the 1,000+ students that pass through the hallowed corridors of the schools. For a few bucks more, a mechanical engineer would trade for a 4X4 cubicle in a software consultancy against a buzzing, full-of-action shop floor 4 times the size of a football ground? There are cases of MEs, who, after their masters from an IIT, have coded in COBOL for the notorious Y2K issue! 

Of course the disparity in pay scales has also shrinked today and there are mechanical engineers who are paid more than their software or management counterparts. That is also because these select few who have chosen to stay in the field have all taken up some kind of higher education (masters and PhD) and they have grown to be so good in their fields that they are close to being irreplaceable.

I remember an old engineer had once said - mechanical engineering is the mother of all branches. Of course, that was at the time when there were barely three streams in engineering - civil, electrical and mechanical. And an odd university offering chemical engineering. These days, even a mediocre engineering college offers at least 10 streams that include unheard of types like transportation, printing, biomedical and information technology (no offence to these engineers, this is the day of specialization). I wonder if these are the reasons behind the dwindling numbers - the generic mechanical engineering has been split into metallurgy, and others? 

No, out of the 110 engineers of my batch, they were all mechanical engineers, only about 10 of them had the passion to follow the field. Today, if I have to hire a mechanical engineer for my team, I have to sieve through literally hundreds or more of resumes and spend many hours of interviewing before I end up with the right one. And all of them have existing jobs. Mechanical engineers, what has our tribe evolved into?

Concurrent Designing

Cost aside, the biggest advantage of having an offshore design center on the globe is the time difference, for example, the headquarters in US and the extended team sitting in Bangalore. There is of course a big "IF" attached to the statement and that is an efficient concurrent engineering process and tools. Actually, more than concurrent, I would say sequential designing would be the efficiency enhancer. The same tools that enable concurrent engineering would be used.

Consider a situation where the idea is generated through R&D in the US center. A project plan is drawn up that consists of 5 modules given to five group of engineers to work on. These 5 modules would be designed in parallel, in their own sandboxes. Within each group, every engineer would have their own sandbox where the subassembies would be created. One of the teams would maintain the master model where the individual engineer would place their subassemblies so the pieces fall in place to form the complete picture. All PDM tools enable concurrent designing, including features like revision control, access control, part number generation, processes for production release and so on. Apart from this, some PDM tools like teamcenter engineering can work on servers located across the globe, synchronizing and updating the data on the servers automatically and mirroring them instantly.

When the engineering teams are located across the globe, the groups can take advantage of the fact that working hours are spread over 16 hours. The engineer can work during his/her working hours and then check in and the engineer on the other side would take up the same design from the mirror server at that location. This kind of relay desiging and baton passing can happen when the design goals are clearly defined in the statement of work, efficient communication channels are in place and the engineers are not too possesive about their designs! Then, let the magic of collaboration begin.

Parametric Modeling

"I am modeling," I would say to my then-girlfriend-now-wife, just to impress her, every time she called. But of course it would all be shattered if I would prefix "parametric" to it.

Now I do not want to extol about the virtues of parametric modeling, we all know it, don't we? Early in my career I took upon parametric modeling very proudly, announcing to everyone who cared to listen about how easy it was to change the form, fit and function by merely playing with the key variables.I discovered the real advantage when I started performing design of experiments.

The objective of DoE is to determine the optimal value of a certain set of variables in a model would effect the desired outcome. For example, if the task on hand is to determine the optimal thickness of a beam that can withstand the maximum forces subject to it. The usual constraints would be the type of material selection, weight of the assembly and cost. A typical DoE would have more such parameters to vary, the width, height, length, supports and so on. If the beam section has 4 standard variants available with the supplier, these values would be input to the analysis program that would calculate the strength of the beam for each material and cross section. Another spreadsheet would calculate the weight and cost of the beam. The parametric modeling software would take in these values and would cycle through all the variations. (Of course some basic automation program is required to feed these values and run the program sequentially). If these values and outcomes are fed into a statistical program, we would get a transfer function, that is, a relationship between the variables fed into the tool and the desired outcome. And from this transfer function, we would get the optimal beam section and material that would yield the maximum strength at minimum cost.

The CAD software these days are all parametric by default, Unigraphics provides some non parametric features, Pro-Engineer and Solidworks by default have predefined variable names for every feature created in them and they provide a nice intuitive interface to program using them. UG also has these features but the menu is not as intuitive. Ansys has APDL and tcl-tk that is widely used to pass variables to the program. Parameters and I have been fast friends and have helped me cover change requests and perform DoEs quickly and efficiently.

Obsessed with Finite Element Analysis

Those were exciting times in 1996 when in the midst of my mechanical engineering at KREC Surathkal, I first heard about Finite Element Methods. It was glorified by my seniors as the cure-all, solve-all package to tackle ME challenges. I jumped the bandwagon and took up the elective and opted to do my final year project in it. After a miserable year of trying to learn FEM (no thanks to our Mathematics prof who handled the subject in the most mathematical way he could), I almost gave up on it! Yet the lure of FEM was so compelling that after my BE, I did a course at IISc (Indian Institute of Science) to clear things up for me and FEM was no longer the labyrinth of integrals.

Those days, it was difficult to ignore the penetration of FEM, I quickly moved from Bosch to GE to solve more challenges using FEM. It was during my initial stint at GE that I was drawn into the maze where FEM was equivalent to engineering! Everywhere I would see aircraft engines, generators, turbines and glossy magazines containing entire models of cars and engines in meshed form. I had arrived. It was a matrix like dream where everything around me was made of nodes and elements. Just a few clicks of the button, a few intelligent selection of elements, solvers, material properties and the world could be analyzed in ANSYS, NASTRAN and so on. In retrospect, (partly thanks to a few eye opening sessions with my engineering manager in GE Schenectady), I had forgotten why we were using FEM, just to find the stresses, displacements and natural frequencies of everything that we designed? Gone were the design handbooks, bending equations, formulae, Petersons stress concentrations! FEA could answer anything.

Actually I was just plain lucky (of course the hardwork also paid off). I was dealing with straight forward assemblies with isotropic material properties, small displacements, everything that spelled linear. The truth is that the world is non-linear. As weeks turned to months, my prjects became more complex and I realized that the FEA software we are using is just solving matrices and it was the brain sitting in front of the monitors that had to do the thinking, interpret the colourful post processing and most importantly identify and ignore singularities and outliers. And to do this, I had to anticipate the the solution. Back to pencil and book, I started with free body diagrams, bending and stress equations, basically refresh my "engineering" skills. "Simplify the problem, nature is not so complicated."

Screw it! - Part Deux

Thanks to all the feedback I got from my first post, my second will explain the how, why and what of threaded fasteners. Of course, this post will not go into the jargon of types of threads, thread angle and pitch, any wikipedia page will give that information. I will explain the practical use of different machine screws, heads and types.

One of the first design choices we run into in an assembly is to use fasteners or to weld the parts together. This may lead us to answer the "why fasteners." Any two parts that serve some purpose when mated, but need to be disassembled in the future for service or maintenance will have to be fastened by screws, else we can just weld them in most cases. Also when two dissimilar materials need to be fastened, screws come into the picture. Circumstances allowing, I will always have two parts welded to each other to provide a strong, permanent bond.

There is a lot of know-how into using threaded fasteners. Do we use it with or without washers? What is the use of machine screws with built in flange-like washers. How do we use self-clinching nuts? In order to reduce assembly time, I have always used self-clinching nuts and screws with built-in washers. In most cases, I use a washer to use my screw on an oversized hole (I sometimes use oversized holes to widen tolerances and reduce manufacturing costs). Of course, washers are inevitable when steel screws are used over softer materials like aluminum and copper.

The drive of the screw is very important. I have always avoided using a slotted screw driver because it is difficult to get a continuous drive and the driver always tends to slip. My first preference would be a hex-socket headed screw because this provides the maximum grip to tighten with an allen key. Second preference would be Philip-headed screws. The tools to tighten these are also widely available when compared to torx or other drive types. And I will try my best to use the same type and size of screw in an entire assembly for obvious reasons like standardization and time saving.