I began at General Electric Corporate Research & Development (CRD) in February, 1982. Working at General Electric was totally different from my experiences at Dikewood. Because Dikewood was a small government contractor, management spent money only with the greatest reluctance when it could not be directly charged to a contract. The president of Dikewood severely reprimanded and nearly fired an employee for buying a microwave oven for the break room without authorization from him. At General Electric, there was almost no limit to money as long as it was applied to an important company project and management believed the end result would increase GE's profits.

The most important project my group was working on when I joined was developing some of the software used to design the prototype for the first General Electric magnet for magnetic resonance (MR) imaging. The first MR magnet design was a large cylinder containing superconducting coils surrounded by liquid helium. The patient was inserted into a hole down the center, like a torpedo going into a torpedo tube. (Some patients who suffer from claustrophobia refuse to be inserted into them. Newer designs are more open.) At the time, General Electric was buying MR magnets from Oxford Instruments, a company in England. By making their own, General Electric would make more profit and have better control over the quantity and quality of the magnets.

To give you an idea of the money involved, our budget for computer expansion for one year was $450,000. We did all of our computing on a VAX minicomputer from Digital Equipment Corporation, connected by simple alphanumeric terminals sharing its single processor. They were great computers for their time, but by today's standards, they were frightfully expensive.  In addition to the basic expenses of the VAX computers themselves, they also required special rooms with air conditioning and air filters. Our one-year expansion budget covered three 450 Mbyte hard drives, a 16 Mbyte memory expansion, improved air conditioning for the computer room (the heat sensor kept turning the computer off on hot days, bringing everyone's computer work to a halt), and some color graphics terminals. Today, that amount of money would buy over 100 top PC's. Each one would be 1,000 times faster and have 100 times more memory than our VAX.

The name "MR imaging" has an interesting origin. To physicists, the effect is called NMR for Nuclear Magnetic Resonance. The "N" is important because it is the nucleus that resonates at a precise frequency depending on the magnetic field. However, the "N" word is forbidden in the marketing department because it conjures images of nuclear weapons and nuclear radiation. Imagine General Electric trying to sell a Nuclear Magnetic Resonance imaging system to a hospital versus a competitor selling a Magnetic Resonance imaging system. It's another instance of public relations versus the plain truth.

While most of the people in my group at CRD were working on projects related to MR magnets, I was hired to work on a contract that my boss, M.V.K. Chari, and another staff member at CRD had won from the Electric Power Research Institute (EPRI). The idea was to create a system that would monitor retaining rings of large turbine generators for incipient failures due to stress corrosion cracks. Retaining rings are cylindrical stainless steel jackets at either end of the rotor of a turbine generator.  They hold the copper windings in place against centrifugal forces. Because the rotor of a turbine generator in the United States spins at 3,600 r.p.m. (or 1,800 r.p.m. for nuclear units), a retaining ring splitting apart is the equivalent of a bomb hitting the generator. It has happened to a few turbine generators, resulting in costly repairs and loss of service. An effective device to catch incipient failures would obviously appeal to the power industry.

I was hired because I had experience running experimental projects.  There were some problems with the configuration of the tests as they had been planned before my arrival. To prepare for the project, the previous team had selected an unused room in the building to perform the tests, while the device that was to capture the test data was attached to the VAX computer in another part of the building. They ran an unshielded pair of wires from one room to the other and tested the setup by recording the voltage of a battery. Because they got a non-zero reading on the VAX computer, they assumed everything was fine, although they had difficulty figuring out why the recorded value was an integer in the thousands rather than something close to 1.5 volts as they expected for a battery.  (The reason was that the data acquisition device converted the input voltage to an integer in the range 0 to 4095.)  The VAX had to be switched to data acquisition mode to gather data, locking out all of the time-sharing terminals connected to it and effectively bringing the work of everyone else to a halt. I have to give them credit at least for knowing that they needed to hire someone with more experience in taking electromagnetic measurements.

After experimenting with several types of electromagnetic sensors to detect a crack in a scale model of a retaining ring, I built a simple prototype sensor consisting of a small permanent magnet with many turns of thin copper wire wound around it.  If a crack passed beneath this sensor, a voltage signal was generated in the coil.  I worked with an assistant who helped me refine this idea to create a complete system based on a PC that would scan the surface of the retaining ring while it rotated and generate a color map on a computer screen showing any detected cracks.  When we applied for a patent, we discovered that nearly the identical system (minus the PC) had been used in World War II to inspect brass casings for artillery shells.

While we didn't get a patent out of the project, it still seemed that the system would work until we talked with the engineers at GE Power Systems Department who designed the generators.  Unfortunately, the region we were trying to monitor (the retaining ring) experiences high electromagnetic fields and the engineers expressed their reluctance to insert any objects that might affect the long-term reliability of the generator.  The second problem with the idea is that stress corrosion cracks tend to start on the inside of the retaining ring, a place inaccessible to sensors of the type we envisioned.  It struck me that we could have saved ourselves time and EPRI money if we had discussed these things with the generator engineers during the proposal process or at the beginning of the project.

When it became clear to me that the EPRI project had no future, I asked my boss (whom we all called Chari) to assign me to another project.  After some understandable reluctance because I was hired specifically for the EPRI project, I was given a challenge: Take the separate software programs that had been written to analyze the magnetic fields from MR magnets and create a more general system that an engineer at GE could use to calculate the magnetic field of a devices he or she is designing.  My colleagues and I were successful, creating the first version of a two-dimensional, low-frequency electromagnetic finite element analysis package, widely used in GE, called GE2D.  (I'll explain these technical terms a little later.)  My job as team leader was to create the data interchange definitions among the various programs comprising GE2D and to coordinate the activities of the contributors so that everything worked together correctly.  I also worked as a technical contributor to the project, adding features to the programs to meet anticipated needs of users.

The original GE2D suite of programs ran on VAX minicomputers with Tektronix color graphics terminals.  The graphical capabilities of the programs seem primitive today, but they got the job done cost-effectively for the GE product departments.  Almost every GE product department had a VAX, and individual Tektronix terminals cost "only" a few thousand dollars.  Another cost incentive was the lack of license fees for GE2D.  Portions of GE2D are still used in GE product departments today, fifteen years after its introduction.

Now, to explain the technical terms of GE2D:  First, low-frequency electromagnetics means problems where the electric and magnetic fields are varying slowly enough that we can accurately ignore the wave propagation of the fields.  Devices that can be analyzed with low-frequency electromagnetics include four important GE products: MR magnets, motors, generators, and circuit breakers.  The term two-dimensional refers to modeling only a cross-section of the device we are trying to analyze, not the complete three-dimensional object.  Two-dimensional models are faster to solve and take less storage on computers than three-dimensional models.  Finally, finite element analysis means that we take the cross-sectional model of the device we are trying to analyze and split it into small triangles and/or quadrilaterals, each one small enough so that the electric and/or magnetic field is nearly constant from one side to the other.  This allows the mathematical equations for the electromagnetic field to be approximated numerically and solved with a computer.

Because of the success of GE2D, I was given a greater challenge by my GE management the next year (1986): to add other engineering disciplines such as heat conduction, stress analysis, and fluid flow.  In this, I was only partially successful.  The difficulty was not technical, but came from the fact that the project was stepping on the toes of other research groups at CRD.  Managers of the other programs used the time-honored method of passive aggression to undermine the project, agreeing with our upper management that such a coordinated project would be a good thing, but refusing to provide technical assistance in their respective areas of expertise when I requested it.  By hiring an RPI post-doc as a consultant, we were able to succeed in spite of the obstacles.  By the end of 1987, we had a complete two-dimensional analysis package that combined electromagnetics with heat conduction and stress analysis.  It allowed sequential coupling among the three disciplines.  For example, the power losses calculated by the electromagnetics portion could serve as sources of heat to the heat conduction portion.  We called the package CAE2D for Computer-Aided Engineering in 2 Dimensions.

Despite the technical success of the project, CAE2D was not a political success and GE corporate-sponsored development stopped after 1987.  The total cost to GE to develop GE2D and CAE2D between 1985 and 1987 was about 1.5 million dollars.  While this amount seems astonishing, you should bear in mind the high overhead rate at CRD.  Overhead is what a customer must pay to obtain the services of a worker in excess of his or her salary.  It includes benefits, office supplies, equipment, advertising, rent or building amortization, taxes, and any other fixed expenses.  At Dikewood, the overhead rate when I left was 88%, meaning that if my salary was $50,000 per year, it would cost a customer of Dikewood $94,000 to hire me for a year.  The overhead rate at CRD was about 285%.  It would cost a customer of GE $192,500 to hire me at the same salary.  The 1.5 million dollars to develop GE2D and CAE2D amounted to a three-person team working full time on the project for two years.  Viewed that way, the cost was reasonable.

About the same time as the end of the CAE2D project, the mission and funding method of CRD began to change.  When I joined GE, CRD's funding came 75% from GE corporate funds and 25% from research contracts with GE businesses, EPRI, the Department of Defense, and other funding agencies.  The 75% covered fundamental and long-range research, while the 25% covered short-term applications.  Around 1987, the CEO of GE, Jack Welch, and the GE board of directors decided that fundamental research was inappropriate for a profit-making corporation.  They wanted the GE businesses to have more control over the direction of research at CRD.  To accomplish this, they changed the funding scheme so that only 25% came from corporate funds while 75% came from research contracts with GE businesses and outside sources.  Long-term research was limited to subjects that would have a direct impact on existing GE products or would create a new line of products.

CAE2D, while not the big success we had hoped, did provide us with an opportunity to survive under the new funding arrangement.  By adding thermal and stress analysis, we had mastered the technique of adding new features to our software package.  We canvassed the GE businesses who were using our basic electromagnetic software and offered to add new analysis methods and features according to their needs.  This internal marketing technique worked well for many years.  Our biggest customers were GE Power Systems (for generator design), GE medical systems (for MR magnet design), and GE Aircraft Engines (for designing military jet engines with low radar cross-sections).  GE aerospace (defense business) also provided funding for electromagnetic analysis techniques for radar and microwave  systems.  Strong and steady support from GE product departments was essential to our group, because our overhead rates were so high that we had difficulty winning external contracts.  At its peak, our group had ten people.  Our yearly salary budget approached two million dollars.  Most of that came from contracts with GE product departments for their specific electromagnetic analysis needs.

Throughout the 1980's and early 1990's, Jack Welch continued his campaign to improve the profitability of GE.  He pushed each GE business unit to be either number one or two in its market, the theory being that major players in a market make the rules and make most of the money.  Businesses below that goal would be fixed, sold, or closed.  Several businesses that might have used the services of our group were sold because they were not profitable (enough).  For example, the GE defense business was sold to Martin Marietta, the GE small appliance business was sold to Black and Decker, and the GE consumer electronics business was sold to Thomson, a multi-national corporation based in France.  It was good for GE stockholders (of which I am one), but not good for the Electromagnetics Program at CRD.

One particularly humorous example was the sale of Utah International, a mineral exploration business owned by GE when I joined.  Utah International used a technique involving low-frequency electromagnetic fields to locate buried ore deposits.  I had a small project to model the exploration process on the computer.  I was literally in the middle of a presentation of my results to Utah International management when a GE middle manager came into the conference room to announce that Utah International had just been sold.  He instructed me to end my presentation immediately, copy all of the computer programs and data files related to the project to a computer tape, send the computer tape to Utah International, and stop all further work.  At least I didn't have to write a final report!

While Jack Welch was downsizing GE, Gena and I were upsizing our family.  Geoffrey was born in December 1983.  Gary Edward was born in April 1989.  Gena was so ill with the third pregnancy (toxemia) that she had to stay in bed for almost a month.  She was fired from her job in the Emergency Department of St. Peter's Hospital in Albany because of the time off due to complications of the pregnancy.  We had to choose between that job or mom and baby's health: no contest!

My boss, Chari, retired from GE in 1992.  I was appointed acting manager when he left, and later I was made manager of the group.  GE sent me to management training class at its world-famous Crotonville, New York, training center.  The training center is a combination of a hotel and conference center.  Seating at the large round tables in the dining room was strictly by order of arrival for dinner so that attendees would mix and begin to establish networking contacts.  The food was very good - equivalent to a four-star restaurant but not quite five-star.  Each wing of the residence area had its own kitchenette with a stocked refrigerator for lighter meals and snacks.  The training program occupied almost all of our waking hours for a week.  We were lucky it was only a week for the first management course.  The second management course (which I never took) lasted four weeks, with only one weekend off in the middle.  I talked with a woman I knew from CRD who was taking the second course and she told me the time away from home was hard on her children.

We learned basic management techniques in the course, covering topics like managing projects with interacting teams from different parts of the world and giving constructive criticism to subordinates.  My favorite part was a video tape starring John Cleese of Monty Python fame.  He humorously demonstrated how not to deal with an employee's personal problems.

I learned many things at the course, but the most important lesson I learned was that I did not wish to pursue a career in management in a large corporation.  I had seen enough of the paperwork and reporting requirements of the next higher level of management at GE to know that I wanted no part of it.  I didn't mind leading a team working toward an objective, but I wanted to spend at least half of my time working as a technical contributor.  At CRD in past years, that was the normal mode of activity for a program (lowest level) manager.  However, as Jack Welch removed management layers to make GE more efficient (in theory), many management and reporting responsibilities devolved from the removed layers to the program managers.  By 1994, most program managers at CRD spent their time marketing and reporting to upper management.  Few did any significant technical work.

At the same time, the contract funding sources for electromagnetics began to dry up.  As pressure to produce profits increased, GE product departments switched their funding of CRD to focus on projects with immediate benefits, such as designing a new product or debugging an existing product.  The "Research" part of "Corporate Research & Development" became a smaller component of the total funding each year.  Many senior staff members, notably Nobel Prize winner Dr. Ivar Giaver, left CRD to pursue research in academic institutions or to take early retirement.

The layers of management above me recognized the funding problem for my group and suggested that we find new project areas to work on.  The one negative thing I will say about my experiences at CRD is that the managers above me were absolutely no help in finding appropriate new projects for us.  Their contacts and influence throughout the company would have been a great resource.  We explained the situation to all of the departments in GE that were using the software we had developed over the years.  They were sympathetic and expressed the hope that the group could continue to work on small projects for them and support the existing software, but they were unable to find sufficient funding to keep the entire Electromagnetics Program running - at least 1.5 million dollars per year.

By the end of 1994, it was clear that the Electromagnetics Program would cease to exist as a separate entity.  Some members transferred to other programs at CRD.  Five of us, including myself, decided to leave GE in early 1995 and start our own company, Skyblue Systems.

Last update: June 07, 2000