COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002				A DISSERTATION PROPOSAL
C00026 ENDMK
C⊗;
			A DISSERTATION PROPOSAL

			December 31, 1973


	The  automated  machine  shop.     Two  possible routes, both
viable, one a lot more ambitious than the other. The  more  immediate
sort  of  automated machine shop- the interactive machine shop.   The
more ambitious type of shop- the completely automated shop.   I  will
discuss my thinking on these two proposal alternatives.

	The interactive machine shop.

	Presently,  most prototype machining is done manually with no
numerically controlled machines being employed.    The  prime  reason
given  is  not economics, but the fact that the machinist must really
be there to make the first part.  N.C.  machine  programming  methods
today  require  an  average  of  3  rounds  of  iteration  to  make a
successful part.  Thus, for short runs, the N.C.  machine-programming
debugging  operation  offsets  the  advantages  of the machine's high
speed and errorless machining ability.

	Generally,  programming of these N.C. machines is not done by
the machine operator, but by a programmer who usually happens  to  be
either  an engineer, or an ex-machinist who has had special training.
The operation works like so.   The shop recieves  a  print  from  the
engineering department.  The foreman looks the print over and decides
how to make the part.  If the quantities are  large  enough,  he  may
decide  that  n.c.   machining may be the best way.  In this case, he
will turn the drawing over to a programmer  who  will  use  either  a
Flexowriter,  or in larger establishments a mini-computer and the APT
programming language to create a program tape for the n.c. machine. A
part  will  be  attempted.  Only in rare cases will the first part be
completely successful, as even the best programmeers have  difficulty
n picking the proper machining speeds for all cuts, or accounting for
part or tool deflection, or insuring that no interferences  occur  in
the  cutting sequence.  In just about every case, the programmer ends
up watching the first and second and thrid part goo thru the  machine
while  making  corrections  to the program. Frequently as many as ten
iterations are required to successfully make a part.

	My first proposal is for a more interactive approach to  this
machining operation, involving more computer use with a standard n.c.
machine.   Here is what I envision.  Think of the  following  layout.
You  have  a  prototype machine shop- execpt it is equipped with n.c.
machine tools (in addition or instead of the  conventional  manuallly
controlled  tools).    A  computer  terminal  from  a large timeshare
computer also is in the shop.  In addition there is a machinist. This
fellow  may  not  be  too  skilled  a programmer, but merely a person
capable of taking instructions and making judgements  about  what  is
good  machining practice, and who also knows what his machines can do
and what they can't do.

		In the full blown version of this interactive shop, a
part gets designed by an engineer  sitting  in  front  of  a  display
console.  Here  he has the ability to manipulate geometric shapes and
call in subroutines of fixed configurations, to create a  part  which
he  then  wishes  to  make.   Maybe  he will design a complete device
consisting of several parts, some  purchased,  some  made,  and  some
modified  pruchased  parts.   Each part can be detailed separately on
the screen, and then dimensioned so that it will fit into  the  whole
system,  using  a computer aided dimensioning system which would have
knowledge of  standard  machining  tolerances,  and  practices.   The
program  would  also  have  knowledge  of  other constraints, such as
dimensions or shapes  of  purchased  components  which  fit  into  th
system.    By  making several passes at the program and reviewing the
finalized assembly and dimensioned drawings, the engineer can operate
interactively  with  the  computer  to  produce a system with all the
proper component tolerances,  dimensions,  and  configurations  which
satisfies him, and also represent a reasonable machining task.

	Next,  the  engineer  "sends"  the layout and drawings to the
machine shop.  This means that the computer makes up a machining tape
per  the design drawings, and supplies a list of required material to
the shop.   Here the machinist has a say of  what  the  shop  has  in
stock, and what kinds of jigs and fixtures to hold the parts the shop
may have.    His input is  back  to  the  computer  in  the  form  of
answering questions put to the shop by the computer.  These questions
await simple replies, such as the stock dimension, or the location of
a  corner  of  the  stock  or  part on the mill table, or the size of
cutters available.  Based on this information and a built in program,
the  computer  creates a first pass n.c. tape. The machinist, who may
have a reference set of drawings at hand so that he can spot  obvious
errors, sets up the machine per the instructions, and then starts the
operation.   Rough cuts are performed and unless the  operator  notes
any  fatal  errors, the machining sequence proceeds with the computer
giving the machine AND the operator new  instructions  from  time  to
time.    Feedback  from the operator is accepted and computer updates
in the machining operation are  made  as  needed.   Typical  feedback
information will be in the form of answers to various questions posed
by the computer- such as cutting speed ok? or can  I  take  a  deeper
cut?,  etc.  The  response  will  be in the form of simple Y and N or
number type answers.   Requests for dimension measurement may come as
the  part  nears  the  finished  dimensions.   Here the computer will
request that the operator measure the thickness of  a  part,  or  the
diameter of a bore, or the taper of an edge.  The data is fed back to
the computer and serves to update the program to accomodate  for  the
particular  errors  inherent  in  the machine, or the setup.  Furthur
instructions such as requesting a statement of the surface finish, or
requesting  that  an  edge  be deburred, or requests that a cutter be
changed, or that a hole be hand tapped, or even that the operator  do
a  particularly  sensitive cut where extra careful observation of the
cutting operation may be required.  These commands would all come  to
the  operator through the remote trminal which could be a teletype or
similar device..

		In a more advanced sort  of  system,  the  data  link
between machine and computer could be improved with the incorporation
of  direct  force  feedback,  so  that  cutter  forces,   or   cutter
temperature  could be fedback to the computer.   Even sounds and some
special purpose sensor outputs could be used to furthur  improve  the
speed and accuracy of the prototype machining operation.

	This  completes  an  outline  of a proposed development of an
interactive prototype and short run machine shop.  I  consider  it  a
very  realistice  type  of  proposal.  Several parts of this proposal
could be implemented quickly right here at Stanford.  I  have  talked
to  some  of the people in the Department of Chemistry, and they have
expressed interest  in  a  project  of  this  sort.   Their  shop  is
presently  just  about the best shop on campus, but it lacks any n.c.
machines. In fact, I doubt that there are any n.c.  machine tools  at
Stanford  University,  with  the  possible  execption  of  SLAC. With
machine shop costs of $12 per hour  and  up,  it  appears  that  this
interactive  approach  is  a viable way to speed up the production of
prototype and short run production parts, at a large savings in cost.
Back  to  the Chemistry Dept. Besides having the best shop on campus,
(machines and facilities), they also have some of  the  most  capable
machinists, one of whom is a trained operator of n.c.  equipment from
a previous postion at Hewlett-Packard.     Because  of  this  quality
shop,  the  Chemistry  Dept.   has  a  very positive attitude towards
acquisition of new equipment and thus there are funds  available  for
the  purchase  of  an  n.c.  milling machine(for example).  Here is a
really good chance for  several  graduate  theses,  in  the  Computer
Science , Mechanical Engineering and Interdepartmental Fields.
		A DISSERTAION PROPOSAL

		December 31, 1973

	A continuation of Prop2.


	The totally automated machine shop.

	The concept of the totally automated machine shop was brought
removed and placed in a shear or saw very easily  by  a  manipulator.
up in Prop1.   This earlier discussion outlined some of  my  thinking
regarding  the  possibilities  of  developing  a completely automated
machine shop in  which  the  operator  would  be  a  computer  and  a
mechanical  arm.   The input would be an interactively created design
and the output a finished part or eventually a  product  manufactured
to the description created by the designer.

	As a starting point, I will give a brief description of how I
see the general prupose shop working.  First lets look at the layout.
Imagine that we have a large timeshare computer, and a high data rate
link to a machine shop.  In this shop are a number  of  machines  all
n.c. type or else powered such that either electrical switching or at
the most low mechanical forces are required to make them perform  all
tasks.   They are all located in known locations, and each has a well
stocked accessory holder.  A stock room with a good supply  of  stock
material  is  near the tools.  This stock is stored so that it can be
Besides   these   machines   there  exists  one  or  more  mechanical
manipulators which can move around to all of the machines and perform
various operational tasks on these machines, just as a human operator
does.      The  machines,  the  manipulators,  andα  the   associated
accessory  and  sensing devices are all interfaced to the computer so
that  they  are  directly  controlled  (thru   a   mini-computer   if
necessary).

	Now  lets  look  at  a  typical  sequence of operation.    An
engineer sits down at a graphics display console.   His  task  is  to
create  a  working  device  of  some sort.   This device is part of a
system.    There are several parts to  the  device,  some  purchased,
others  made  from purchased stock, and still others modifications of
purchased completed components.   These parts must all  fit  together
and  operate as a device, and the device must operate properly in the
completed  system.         Using  GEOMED  or  a  similar  interactive
display-graphics program, the engineer can reate shapes and structure
on his console and manipulate these developed  bodies  at  will.   As
engineer  are  generally rather realism oriented people, the graphics
program will probably be oriented to display shapes and  forms  in  a
manner  which  is  easily  interpreted  by  the design engineer.    A
library of pruchased component dimensions  and  specifications  along
with  standard  engineering dimensions and details can also be called
by the designer or directly by the computer, as  would  be  the  case
when placing screw holes, or selecting screw or shaft sizes.

	Assume that the design is done, the designer looks over  what
he has created on the screen and sees that everything is correct; the
program has picked the proper screws, has matched  up  all  the  bolt
patterns  properly,  and  has  chosen  the  right  tolerances.  Now a
manufacturing operations program can be called in.   This  gives  the
operator  a  list  of the proposed machining operations and a list of
the stock required, and the estimated time  and  cost.   A  breakdown
gives  the  details,  so  that  the designer can change dimensions or
specifications to reduce costs, or get around difficult or impossible
operations.  Once this iterative operation is completed, the required
setup list and stock list is compared with the shop inventory.     If
they are complete, the machining can begin.

	Following  the  computer  generated  sequence  commands   the
machining  operations  are  performed.    The individual machines are
directly controlled by  the  computer  using  position  feedback  and
feedback  from  permanently  mounted  motor  current  and temperature
sensors, force sensors,  etc.   Setting  up  each  machine,  changing
tools,  and  transferring  material  are the mechanical manipulators.
These arms serve the same  purpose  as  the  human  operator  in  the
typical  n.c.   machine shop.  In addition, they provide the computer
with a device for  positioning  measuring  instruments,  checking  on
surface  finishes, and making all the necessary observations that are
required of a human macchine  operator.    This  information  is  fed
directly  back  the computer for updating of the machining sequences.
In this way, the shop need not be a very  precisely  set  up  layout.
Real  machines  setup almost casually can be used in such a situation
where feedback is sufficient.

	The task is complete when the finished parts are delivered to
the  output  box,  just  like  line  printer output.   Possibly, even
assembled into  a  complete  assembly,  properly  inspected,  tested,
documented and certified.

	Certainly,  the  development  and  execution  of  a complete,
general purpose  system  such  as  has  just  been  described  is  no
overnight  task.    A number of man years effort is involved, both in
the programming and the engineering of such a system.  The  execution
of  such  a  problem  is  beyond  the  cope  of  a  single PhD thesis
dissertation, but a thorough study of the problems  and  approach  to
such  a  project  may  very  well  be a good thesis topic. But, as an
alternative to a paper study of such  a  general  purpose  completely
automatic system, it seems reasonable to attempt the development of a
special purpose automatic shop as  a  more  realistic  initial  goal.
What  follows  are  some  thoughts  on  a  special purpose completely
automated shop.


		The automated sheet metal shop.

	Prototype sheet metal parts are very  expensive  relative  to
production  quantities.    As an example, it is frequent to find that
the cost of just two of a kind is only 5% more than the  initial  one
piece. Why is this so.  Well, there has never been much automating of
sheetmetal processes, other than  blanking  and  stamping  which  are
restricted  to  large  production  runs.  Other than n.c. punches and
n.c. stops on hand fed shears, there are no  other  really  automatic
machines  used  in  this  field.   It  has  been considered a hard to
automate field, because of the types of machines used and the need to
do  a  lot  of  manipulating  of the material which can frequently be
large floppy sheets, of  varying  thicknesses,  yield  strength,  and
stock dimension.