That's what Brian Carlisle says people still typically have to endure nowadays when they are installing automation. While anyone who has been involved in an automation project probably agrees with that assessment—and also with his points that there tend to be perceived risks associated with automation performance and costs, and that the one-promised capital and engineering utilization of flexible automation, in particular, seems to be more talked about than realized—there's something peculiar here.
You see, Brian Carlisle is the co-founder (1983) and present chairman and CEO of Adept Technology, Inc. (San Jose, CA), one of the leading automation hardware and software companies in the world. There are, for example, more than 15,000 Adept robotic systems installed world wide, including at many automotive OEM and supplier companies. It seems more than odd to sit in Carlisle's office and hear him be quite so methodical about some of the things often not talked about by automation vendors.
Carlisle isn't blaming anyone with the remark about lateness. Or about the issue of low (or no) reutilization of capital and engineering (how often are feeders, tools, and software tossed when going from one program to another?). Rather, Carlisle is simply being candid about a state of affairs that he maintains is untenable.
Too Much. Too Late.
He sees other problems, too. Like the issue of capacity. He notes that there is a tendency to estimate the amount of capacity that will be needed for a program—say something on the order of 100,000 units per year. But the user may only hit that volume a couple of times over time during the ramp up and ramp down (assuming, of course, that they actually hit 100,000: otherwise, they've just paid a whole lot for a system that won't hit its established capability—and if the aforementioned issue of lack of reutilization is valid, then it isreally expensive).
And as for the ramp up: Carlisle is also concerned that there is no science base for analyzing system yields, so consequently, there are long qualification times.
Carlisle and his people have been doing a lot of work of late with people making consumer electronics and cell phones. He points out that in the case of a product in these areas, the product may have a nine-month life cycle, and the ramp up period for the automation program may take six months. This is far from being an ideal circumstance.
So what is to be done? Although Adept is and will continue to be in the robot arm business, Carlisle admits, "As we look at the business, we recognize that customers don't want to buy SCARAs, they want an assembly system." He amplifies: "Robots are not the issue. Systems integration is."
In other words, Carlisle maintains that the problem with automation has little to do with the reliability or the performance of robot arms. People want to produce parts automatically (or semi-automatically). Part production requires a system, which may include robots—SCARA arms, rectilinear arms, whatever. But if a system takes six months to get up and running, and the life-cycle of a product is nine months, then there is a whole lot of investment not being optimized.
|Offered as an alternative to bowl feeders, the FlexFeeder system utilizes a vision-guided robot to pick up parts. When part types change, all that's typically needed is a program change. This is the sort of capital reusability that Carlisle thinks is important for automation installations.|
What Carlisle and his colleagues are working very hard is what they are calling "Rapid Deployment Automation" (RDA). The objective is simple: to reduce the amount of time, risk and cost associated with the implementation of an automated system.
This is not a situation wherein there is a cadre of Adept people who will go rushing into a site to quickly install a system. In fact, one of the approaches that Adept uses is to let systems integrator companies do the systems integration rather than having its own people do the task.
Rather, Adept personnel are developing what can be thought of as a toolkit, one containing both hardware and software, that will allow the design, engineering, setup, installation, ramp-up, and ramp-down of a system, as well as reutilization of system components, in a fast, systematic manner.
Specifically, Carlisle is talking about such things as having a 10-station flexible automation system up and running in eight weeks, with six-sigma performance at each station within two weeks. When it is time for changeover, there will be as much as 90% reutilization of capital. This is what RDA is about.
On the one hand, DFFA includes things like the company's FlexFeeder 250 parts feeding system. Sal Brogna, Adept's director, Robot Systems, developed the FlexFeeder in 1993. He explains, "We were trying to come up with a generalized solution for small parts feeding [i.e., parts that weigh no more than 2 oz. and which have a major feature dimension of not more than 2.5 in.]. We didn't want to have custom engineering for each new part that would be handled." Essentially, the FlexFeeder 250 consists of a two-tier conveyor setup. Parts fall from the conveyor on the top to the one below. The falling helps separate the parts. A four-axis, table top SCARA robot is guided to parts based on input from a vision system, the AdeptVision VXL. The pieces that don't get picked within the cycle time fall into a bin at the front of the unit and are recycled back for another pick cycle.
"Bowl feeders," Brogna says, "are good for high volumes of the same thing." He mentions that they tend to jam, but he lets that point go. "The second that go you into variables..." It can take, he suggests, from six to nine months to get a bowl feeder engineered, so part changes can drive time and cost, and fill up physical space on the factory floor. With the FlexFeeder approach, it is mainly a matter of reprogramming to handle different parts.
A second-generation FlexFeeder is being developed by Brogna and his staff in something of a "skunkworks" that Adept has setup in Livermore, California. This system has what's described as "active reorientation." In effect, individual parts on the lower conveyor are tapped from below with a defined amount of force that causes the part to flip into the position that is optimal for the robot to pick up. This further enhances the speed of part picking.
This modeling of the part's physical characteristics is part and parcel of another important aspect of the RDA methodology. Carlisle explains that what's needed by industry is a better understanding of the entire system, including the part models. For example, with regard to part picking, he says that kinematic constraints and the coefficient of friction must be better understood so that there can be a better match between the part and the gripper, thereby reducing the opportunity for error. Math models can replace the conventional trial-and-error.
|The Digital Workcell package allows users to start programming the system prior to the physical build of any of the elements. This facilitates concurrent engineering and leads to faster product launches.|
In 1995 Adept Technology acquired a simulation company, SILMA, which is presently developing many of the software simulation tools that will facilitate Carlisle's vision of RDA. One way of thinking about where simulation fits is to consider it as the slash between CAD/CAM. For example, it has an off-line programming and simulation package for coordinate measuring machines called CimStation Inspection (which was selected by Ford last year as the standard software for these functions; it signed a multi-year agreement valued at $15-million). It works with 3D CAD models (APIs [Application Programmer Interfaces] exist for the major CAD software packages) and kinematically accurate models of an array of coordinate measuring machines and probes. The result, Timothy E. Coogan, SILMA's business director, says, is the capability of performing accurate off-line programming "five to 10 times faster than conventional off-line systems" Coogan says that in addition to Ford, CimStation Inspection is being purchased and evaluated by several automotive companies.
With regard to robots, Adept has launched a new 3D robotic assembly simulation product, Digital Workcell, which allows programming for a robotic cell to be initiated prior to the building of any of the hardware. Cycle times, program logic errors, unreachable location errors, collision errors, and motor saturation errors can all be identified early on in the cell development process, thereby helping reduce the required launch time.
"Even for automotive companies, the batches are getting smaller," Carlisle says. His objective with RDA is to make making changes to flexible automation systems as simple, predictable, scalable, and reliable as possible, and to help users reuse the hardware and software models to the greatest extent possible. By doing so, he thinks that the overall utilization of flexible automation will increase, which is, of course, good business for Adept Technology.