Industrial Robotics: Setting Expectations for Success

The saying "I am from Missouri, show me" describes the requirements often demanded by a firm that is contemplating the acquisition of a robotic system. The system was conceived to deliver a specific set of performance criteria, and the user wants simply to validate that the system has the ability to fulfill the requirements. This requirement is the essence of the definition of validation. For a mass-produced unit like a machine tool or a welding power supply, validation is straightforward. Machine prints enable manufacturers to continue to build machines to a known specification with known capabilities. All the inputs that go into building the machine, as well as the performance criteria of the machine, are constants. The objective for builders of standard robot systems is to meet the user's requirements without having to validate anything relative to the system. The manufacturing world loves standards, and standard systems enable suppliers and customers to agree on the contract requirements without validation. The fulfilling requirements of the system are known to all.

Robotic standards have proliferated across the robotic arc welding segment more than any other application segment. Standards are widely accepted within the welding segment because there are only a few variables to manage in fulfilling the process requirements.

FIG. 1 illustrates four categories of basic standard welding cells, offered by Genesis Systems Group, that make up the primary offering of standard welding system configurations used in the marketplace. For the welding example in FIG. 1, the user needs only to match the system relative to the work-piece weight and size, and decide whether rotation of the work-piece is required to maintain all the welding in the flat or horizontal position. Simple choices are made, and in minutes the system is selected. The user then needs to consider the work-piece fixturing requirements in terms of weight and size in addition to the work-piece itself.

====

Generic Configurations

Fixed Tables

* Headstock/tailstock based systems

'Turntable systems H frame and Ferris Wheel system

FIG. 1 The World of Standard "Platform" Robotic Arc Welding Systems

====

Configurable within the standard platform:

* Robot size and quantity of robots

* Positioner type, size, payload, swing diameter, tooling length, and quantity

* Type of welding power supply

* Robot riser height

* Control system ( through the robot teach pendant, or external device )

Tailstock ( fixed or sliding, stroke )

* Power requirements ( ZZOVAC, 48OVAC, etc..)

* Optional items like welding torch cleaners, other

* Mechanism to change fixtures out of the weld stations Variables remaining to enable the platform to be placed into production:

* The welding fixture to be used (manual, or automatic clamping)

* Welding procedures

* Robot programming FIG. 2 Variables and Configurations for a Standard Robotic Arc System

====

FIG. 2 shows a stand alone cell that embodies the configuration changes that can be made within the standard system, and the variables that need to be managed. The definition of a standard robot system is an integrated set of modular components on a common platform. Each of the components or modules can be added and expanded to enhance functionality for handling larger-sized products, and robots can be added to increase throughput. Welding stations (tables and/or positioners) may also be added.

Once the standard system configuration is selected, the standard platform requires planning of installation, programming, and applying a welding fixture before an order is released to production on the shop floor. This robot welding platform configuration example is also usable for other application platforms in terms of an expandable set of integrated modular components on a common base.

Limitations, regardless of application type, are work-piece size, process complexity, and flexibility. When the work-piece size exceeds the standard system capability, then a more custom-engineered system, still using modular components, will be necessary.

When the process complexity or flexibility exceed the capability of the standard system, a more conventionally engineered system, specific to the application, is required.

How does one prove that the robot system will work? It's not enough to bring utilities to a standard platform to validate that good product will be automatically produced. In the welding example shown in FIG. 2, the work-piece requires a unique welding fixture. A unique welding program, welding sequence, and welding procedure, are also needed to make a good part. This is where the rubber meets the road on any robotic system, and the same can be said for any robot application. Ultimately the robot must be applied to make a "good" part. How is that statement to be validated? An example in the previous Section described how a standard welding platform struggled to make a good part because of the uncontrolled raw material tolerances in the pre-welded condition, and as conditions changed while welding.

What if the fixture could not position two components within the tolerance required to make a good part. What happens if the programming time exceeds the expected robot weld cycle time by 30 percent? What if a good weld simply cannot be made because of the pre-welded part design? These questions are relevant for any standard platform, and it’s easy to imagine the questions that would arise for a custom-engineered robotic system that wasn't built from a previously- generated set of mechanical and electrical prints.

Validation up front is everything. It takes time and money, but the time and money spent after the system is on the production floor, and not working because assumptions were made instead of validation, cost much more. Higher costs are especially likely for process-intensive applications such as material removal, and welding, where cosmetic expectations (surface finish, bead appearance, visual flaws) of the finished product always define a good or a bad part.

The validation phase in robotic implementation is critical. In welding and material removal, the criteria for a good or bad part is also generally subjective, which makes validation up-front even more important. How do you know you can make a good or bad part without building a solution, and testing that solution within a production-like environment.

Simulation and offline programming tools are wonderful tools but they don't make parts. To fulfill the system requirements, validation of the process prior to building a system, is of vital importance. Good ways to validate the design are to see the same part robotically processed within an existing system, or to develop a proof of concept incorporating tools necessary to demonstrate that the process will make a good part. The proof of concept section of this Section will cover several real examples of where proof of concept was the right step in implementation of the robot project.

The first thing the team should do when looking at matching a robot solution to a project need is to first check how a standard robot platform can satisfy the need. The team will very quickly learn whether the solution fits a standard system when they begin to ask all the questions about how each step within the sequence of operations will work, and will contribute to making a good part. Of course, the solution also has to achieve a specific budget and a specific throughput to be justified, or else it's back to the white board.

The validation step takes all the information that was collected during the audit and discovery phases, and allows the team to examine the information within the world of constraints that the organization needs to be successful.

===

Arc Welding Material Removal Palletizing/De-palletizing Machine load/unload Press tending Throughput (rate), bead appearance, weld quality ( i.e. penetration, tensile/yield strength), part appearance related to weld spatter, sometimes dimensional tolerance requirements, weld defects (i.e. lack of fill, porosity, undercut. weld size, etc ....), welding torch access, orientation, and reach to optimize weld conditions, obstacles in the way, robot reach, as-presented raw material tolerances Throughput (rate), surface condition and defects (ie., scalloping, marks, reflectivity, finish, etc..), dimensional tolerance to parent material, cleanliness, robot reach, access to work-piece surface area to be processed, obstacles in the way, orientation of tool or part to optimize process, robot reach, as-presented raw material tolerances Throughput (rate), unit load pallet pattern, robot payload carrying capacity, robot reach, obstacles in the way, securing product adequately in gripper throughout the pick and place cycle, pick and place to1erances of product, product mix by size Throughput (rate), part loading tolerance ( location and radially ) into work-holding, robot reach, obstacles in the way, robot payload carrying capacity, gripping method to pick and place work-piece within process tolerances. gripping forces on work-piece, locating part at inbound relative to loading tolerance in machine, number of robot axis, product mix, type of machine and work-holding Throughput (rate), tolerance for locating the blank in the brake prior to forming, ability to grip surfaces of work-piece through the forming process, obstacles in the way, robot reach, payload carrying capacity, product mix, type of brake

FIG. 3 Validation Process Requirements for Various Robotic Applications

===

Tools for Validating the Robotic Process

Welding and material removal have been discussed from a validation aspect. For non-process-intensive applications such as robotic pick and place, and material handling applications, the team needs to be concerned with other issues. There is a core set of variables that come into play any time a team is trying to develop a robot solution, The development process starts with the parts and the sequence of operations needed to produce the solution that makes a good part. FIG. 3 breaks out the validation criteria for each type of application but in general the solution for any application requires validation of the following list: How to achieve the required throughput

How to manage process variables to produce a finished product that is defined by a set of constants. In other words, how to manage the stack-up of tolerances to make a good part every time How the system makes the right decisions at each step in the operation What risk is being introduced at each step of the operation, based on the team's choice about decision making The world of constraints includes the budget, the condition of the raw material and other inputs into the system, the available space (floor and ceiling), and the sequence of operations in terms of system flow and the flow within the value stream.

Now the team knows the task at hand. It's time to develop a solution that answers the questions above, and falls within the manufacturing constraints that were defined by the buying firm. One of the first exercises to be accomplished is to determine the process and flow that achieves the throughput requirements. Immediately following is the selection of the robot models and the quantity of robots to achieve the throughput. At every step of the sequence, pros and cons of floor space, level of risk, safety, and investment are planned. Many projects fail to move forward early in the audit and discovery phase because what the user wants to achieve fails to satisfy restraints on space, risk, safety, and/or investment. In answering whether a project is good or bad, one has only to look at how the solution matches up to risk, investment, safety, and space.

System concepts often have to be revised, and each revision includes trade-offs where compromising allows for a successful solution but with realistic conditions. Perhaps in the whole plan of robotic implementation, this exercise is the most important in the chain of events. The process and flow must be planned, then the system must be sized correctly for the required throughput.

Resources have to be determined at each step, and the impact on safety, risks, space, and investment are finally examined. A lot of the tools, such as offline programming simulations, and modeling software, allow for quick responses in recreating and validation of new solution revisions. Essentially, this is the world of system integrators and users where, at the end of a quoting process, both parties arrive at a compromising solution. Unless a system fits the standard configuration, the time needed to derive a custom-engineered robotic system can take a few days to months. The amount of time is irrelevant without a cohesive team consensus, so plenty of time is justified in this phase.

Industrial robots are selected by their payload capacity, speed, and number of axes of motion (4, 5, or 6). Also, the environment, such as a wash-down cleansing environment, or hazardous corrosive environment, will influence selection of a specific robot model.

The payload-carrying capacity of the robot defines the maximum weight the robot can carry, including the gripper, throughout its motion range. Additionally, the robot payload must be validated in terms of the moments of inertia, as a function of the center of gravity of the gripper and the work-piece on the various robot axes.

Reach is validated by a reach study using a PC-based, offline, robot simulation software program. Additionally, the team will need to model the system in the virtual environment and begin to jog the robot to the various location positions, with the robot axis in the appropriate orientation, holding a tool and work-piece that are representative of the tool used in the actual process.

The tool, work-piece, machine, or peripheral device with which the robot is interacting, also need to be orientated correctly, as would be required in the production world. In other words, the simulation would incorporate a robot tool such as a vacuum gripper for press tending or palletizing, a cup-grinding tool for grinding, or a welding torch for welding. The simulation also incorporates solid-object models that represent every device and thing within the system, again to validate the process as close to production conditions as possible. Simulating the robot and tool motion throughout the sequence of operations allows the team to validate reach, speed, obstacles, and cycle time at each step in the sequence. The result is the correct selection of robots, and spacing for the various components relative to each other, including elevation heights.

The simulation exercise is a must on every job, including standard systems, because the success of each application hinges on making a good part. The simulation easily allows for quick changes needed to optimize the solution to fit the project constraints. FIG. 4 shows examples of a simulation tool used in the validation phase of a project for welding. The illustration shows an invert-mounted welding robot with a welding torch positioned at various weld locations, and with the work-piece orientated in varying positions in space. The simulation tool will save an immense amount of time by finding problems with part fixture design, robot positioning relative to the work-piece, and other deficiencies such as cycle time.

The simulation won’t manage stacked-up tolerances or validate whether a good part was made. However, in many situations, the simulation is all that is needed to complete the validation process.

The simulation tools also allow for robot motion and entire robot programs to be created off-line on a conventional personal computer, and later downloaded to the robot system to execute the created routines. With every simulation tool, programs created off-line will need to be touched up because calibration between the real and virtual world will have some unpredictable chance of error. The benefit of simulation is that it allows products to be programmed in a PC environment without interrupting the production system.

Programming new products off-line allows a much faster to-market response for the user than if a new product was released directly to the work-cell and a new program was required to be written by the operator each time.

Two primary simulation packages are commonly used. One is the robot simulation software previously discussed, where robots are programmed off line and the software is used as a development and validation tool as well as a programming tool for new products. The other style of simulation software is based on production flow for the entire manufacturing value stream, where the robot content is one entity affecting flow. These tools are not covered in this text but are most useful for the discussions in Sections 6 and 7, where the user is planning on implementing an entire manufacturing value stream from raw material to finished product.

FIG. 4 Example of PC Based Robot Simulation Software Performing a Reach Study for Robotic Arc Welding

The ability to achieve throughput is at the center of most projects.

At the end of some defined amount of time, a specific quantity of acceptable finished parts needs to be completed. The throughput is almost always defined in units of parts per hour, and in palletizing, the terminology is in units of cases, pails, or bags per minute. It’s important to incorporate an efficiency factor in calculating throughput. For instance, in a CNC machining process study, a factor for efficiency is applied to the calculated cumulative cutting time to determine the number of machines necessary to achieve the throughput, because the CNC machine won’t be cutting parts 100 percent of the time.

For arc welding, a similar process cycle time study is applied using a weld calculator. Validating throughput is the foundation that the system content is built upon. For instance, the robot quantity, process flow, how many hours per day the system needs to run (which also drives supporting resources), quantity of process equipment, and the size of the equipment system investment, are all a function of throughput. Throughput also affects the overall size of the system, as does its fitting within the allowable floor space constraints, and the logistics of feeding the system with raw material and removing finished product. FIG. 5 illustrates a Genesis Systems robotic welding process calculator, designed to determine overall process cycle time to completely weld an assembly. FIG. 5.1 illustrates an example commodity for which use of the calculator to determine the throughput for welding a work-piece will be discussed and the conclusions reviewed.

Developed by Genesis Systems Group, this welding calculator requires the user to enter data in the yellow boxes. This data includes the number of inches of weld for each weld size. An associated welding travel speed is given for each weld size. A welding time is automatically calculated as a function of the user-entered inches of weldment for a given welding size. A nominal welding travel speed from the table shown in FIG. 5 is included in the calculation. The arc time is simply travel speed (i/min) / inch of weld (in). This table is a good example of how to calculate overall cycle time because it includes input from the user about non-arc (non-value) tasks, including positioning time, torch cleaning time, touch sensing time, and number of arc starts for the part, and incorporates the air moves of the robot in positioning the weld torch from one weld to the next. There is also a miscellaneous box that should include any load/unload time for placing a work-piece in and out of a fixture. The calculator has no factor for other weld joint configurations such as V- or J-groove styles.

===

79 %" welds 424 inches of weldment 27 3/16" welds 90 inches of weldment 18 3/8 welds 45 inches of weldment Summary: Positioner required to rotate frame to keep the welds in the flat and horizontal position. Assume about a 1/3 of all the welds will be touch sensed 559 total inches of weld, single pass 0.052" wire diameter mixed gas shielding 124 welds ( arc starts ) Throughput summary using rough cycle time calculator Total arc time is 38.7 minutes Total non arc time ( non value added ) is 3.2 minutes With an 85% efficiency factor the total time to weld the frame with a single robot would be 49.3 minutes

FIG. 5 Weld Calculator Example-Frame Assembly

===

Another important factor to consider is that no machine or process operates at 100 percent efficiency, and for our example in FIG. 5.1, a nominal 85 percent efficiency was applied. There will always be some form of downtime related to changeover and setup, non-process time, maintenance, or a device that is not working, etc. If a manufacturing process cannot tolerate even the slightest level of downtime, then redundancy must be incorporated in the manufacturing value stream at all critical steps. The conclusion for the frame example is that a frame can be easily welded using a single robot within a standard system having two welding stations, so that while the robot is welding on one station, the operator is unloading a finished frame and loading/hacking a new, pre-welded frame at the other.

If the customer can accept a completed part every 49 minutes then everything is great. The supplier will con figure the standard system for the right-sized robot, welding equipment, and positioner to weld the frame. The user will supply a holding fixture to hold the tacked assembly, and all that is left is to program the position of the frame at each welding station. Weld paths can be programmed offline and then touched up on the real system, and additional fixtures can be applied in the future, so long as the size and weight are compatible with the standard system specifications. On the other hand, if the user needs a frame every 10 minutes then a solution is necessary that is way beyond a standard system. Factors to exploit include the non-value added time, the efficiency of 85 percent, welding speeds, and the manufacturing flow. We'll come back to this example in Section 7 and review the user's choices to get to a 10-minute cycle time per part to achieve the required throughput.

In trying to minimize cost and maximize throughput, a lot of thought is devoted to validating whether the production rate can be made, using the smallest number of robots and other resources as possible. Examples of drivers that affect the throughput for various applications are as follows:

Application Criteria that affect throughput

Arc Welding Number of welds, and time to access each weld, given the level of obstructions that the robot has to maneuver around.

Welding procedure; number of weld passes, and weld travel speed needed to achieve the proper weld characteristics Amount of non-arc time including robot motion in space, torch cleaning, and positioning the work-piece in the flat or horizontal position for welding.

Load-unload (LUL) time of raw and finished product. Multiple stations allow load/unload time to be within the welding cycle.

Additional process time such as weaving, slowing travel speed to bridge gaps, and searching the weld profile and location because of inconsistent raw material tolerances. The rule of thumb is that tolerances need to be within plus or minus half the weld wire diameter, or robotic technology such as arc-seam tracking, touch sensing, and vision/laser guidance, may be necessary to make a good part See FIG. 5 for a sample weld calculator that determines the approximate robot weld time. Note that an operating efficiency is still required to be applied to the cycle time.

Machine LUL

Machine door open to machine door close time, where the robot is unloading and loading finished and raw parts to the machine tool work-holding fixture. To minimize this time, robots are fitted with dual grippers to speed up the part swap with the work-holding fixture. This time is applicable to lathes but not to machining centers with pallet changers, which enable the robot load/unload time to be included within the machining cycle.

Qualifying time to enable incoming work-pieces with inconsistent tolerances (location, orientation) to a consistent location and orientation for loading the work-holding. Applying a fix ture to incoming product is a way to eliminate the time needed to qualify the locations of the raw in-bound product. The most random form of in-bound raw material is a container where product is randomly located, Using 3-dimensional vision technology will enable the robot to locate, orient, and pick up the product within acceptable work-holding tolerances, although sacrifices in cycle time will be needed.

Product changeover. The robot and robot peripherals can be designed for automatic changeover. The machine work-holding fixture can use redundant tooling and quick-change tooling.

The changeover time will affect throughput because the machine is not cutting parts during this event unless the following changeover features are used.

The workpiece can be re-gripped between machining operations in order to change its orientation for the next machining operation. For manipulating shaft parts, and parts that the robot can grip on the body, the wrist axis of the robot is capable of turning the work-piece 180 degrees so that the work-piece can be held at either end during machining. This feature eliminates the requirement to place the part into a re-grip stand and then re-grip the part.

Ancillary robot motion between one sequence and the next increases overall cycle time.

Palletizing:

Palletizing can control the location at which the inbound product is picked up and the location where the outbound product is placed. Additionally, palletizing can control the number of tasks that the robot must perform in each cycle when palletizing or de-palletizing.

The slowest axis of industrial robots is the waist, so that minimizing the waist axis motion is important. Vertical movement of incoming product relative to unit load height also affects robot motion. Minimizing robot motion to pick-and-place product from one location to another is obviously important.

Palletizing patterns affect the efficiency of each robot pick and place cycle. Each product, on each layer, of a palletized unit load has a specific location and orientation. A simple pattern is columnar, and as the pattern changes from a simple columnar pattern, additional time is required to pick and place product.

See FIG. 6.

Picking slip sheets, pallets, and tier sheets reduce robot throughput because the robot is handling these items as well as the product in any given cycle. There are automatic dispensers of sheets and pallets that will increase throughput.

To expedite handling rates in palletizing, robot grippers are designed to handle multiples of product at a time, up to entire layers of product.

Time is required to transfer full pallet loads out of the robot palletizing zone, and to transfer empty pallets into position.

Conveyors can be designed to expedite this downtime or the robot has another pallet location to palletize to, during the pal let transfer at the first station.

===

FIG. 6 Example of Palletizing Pattern Effect on Throughput

===

De-palletizing product, meaning removing product from a pal let unit load and placing the product in another location, is affected by searching for the product location on the unit load.

Searches might be for partial layers of product, or for inconsistent product location on the unit load. In the first example, the robot must search for the layer at the first pick-place cycle unless an operator signals the robot where to start picking up product. The second example requires the robot to take some additional time to find the product at every cycle. FIG. 7 represents a de-palletizing project where the robot uses a vacuum tool to pick up individual glass sheets and place them on a flat conveyor. If the glass on the conveyor is required to be placed within a tolerance smaller than the tolerance of the glass location on the pallet, then additional search time is necessary to match the tolerance difference at the pick and place locations.

====

There is a required location to place each piece of glass on conveyor I Glass on rack. Stack up of tolerances: The indexing table The pallet location on the table Each piece of glass in the unit load on the pallet

FIG. 7 Example of Tolerance Effect on Throughput

====

Material Removal:

The need to qualify inbound raw material tolerances regardless of the process of taking the media to the work-piece or the work-piece to the media, decreases throughput.

Depending on the tightness of the tolerance requirements for the material removal process, consideration for part presentation is critical. Even with the technology of active and passive force-compliance tooling that enables tools to maintain a constant force throughout the removal process, to cater for variable surface tolerances, the part still requires to be qualified.

Loosely-located product may require Renishaw probes, laser-based sensors, or vision to qualify surface features prior to removal The rate of removing material (feed-rate) is a feature that is critical to throughput. Feed-rate is a function of the media and base material type, but also affects the amount of material to be removed.

Predicting throughput for material removal is similar to a machining time study. Each tool that is used has a time associated with the tool change and the material removal process.

Press Tending:

Separating inbound blanks that are stuck together with oil, and rust, and detecting blanks that are stuck together, is time consuming. Magnetic blank separators, and double blank detectors on the robot tool, may reduce the time taken for these problems.

Qualifying the blank takes away from forming time, but is a necessary step unless inbound blanks are already presented in a qualified fashion.

Time taken to re-orient the part during the forming process.

Tool changing on the press brake is critical, though the robot is capable of changing tools. Brakes allow for multiple tool setups, acting in a sense as a progressive die. Manual tool change will take away from throughput.

If the robot uses magnetic grippers to grip the work-piece, additional time will be required to release the work-piece, often using compressed air.

Many factors affect throughput for a given application. The underlying theme for all the application types is that there are tasks that increase throughput and tasks that reduce throughput. Lean manufacturing tells us to eliminate waste and non-value added tasks. The exercise of validating throughput should force the engineer to ensure that there are minimal waste and non- value added tasks. For instance, welding travel speeds should be maximized by increasing the deposition rate of deposited weld metal, for instance, by using a larger wire diameter, or adopting tandem MIG (dual) welding torches on one robot. Increased throughput can be gained by using the right tools and processes. Another example, which will be covered extensively in Section 6, is the use of robotic vision to replace conventional robot touch sensing, using the welding wire to define weld locations. The speed of this non-value added task can be increased significantly by using vision instead of the welding wire to locate the weld path. In welding, loading and unloading time should be kept within the process time whenever possible, because loading and unloading parts by means of the robot system is non-value added activity.

Robotic touch sensing controls where a weld seam is located, relative to where the seam should be. In other words, the touch sensing routine determines the offset between the "golden" part and the actual part. Tools like touch sensing are necessary with weld seam tolerances that exceed half the width of the welding wire diameter.

The reality is that raw materials (inputs) presented to a robot welding system will rarely achieve the tolerances that are ideal for robotic welding. For any robotic system, regardless of the application, there will be a mismatch between the tolerances of the inputs and the tolerance expectations of the finished products (outputs). The challenge for the engineer is to manage tolerance mismatches while minimizing waste and non-value added tasks, all in the pursuit of throughput. As may be recalled from Section l, throughput is very important in contributing to productivity, capacity gains, and lowering costs.

====

Example 1. Robotic tending of turret punch. 8' x 5' mild steel blanks are presented in a stack of (80) pieces Inputs (tolerances) I; Output (tolerances) on a pallet. The robot picks a single blank from a stack and loads the turret punch table Blank location at the inbound stack is +/-.375" Blank length and width as sheared is +/- 0.062" along the width and length Blank flatness user can't state exactly, but within

+/- 0.060" Blank thickness can L vary +/-.035" Origin location for placing blank Blanks stick together I due to oily surface Blank to be located on the table +I0.035 Blank thickness needs to be within +/- 0.025 Solution to manage the tolerance variance between inputs and outputs:

* Double blank detection in robot gripper

* Requires an inspection station prior to loading the punch using laser guided sensors to capture blank location

* Inspection station also equipped with additional sensor to pass or fail the blank thickness specification FIG. 8.1 Stack up of Tolerances

====

Wheel wear causes wheel diameter to become smaller but the wear is not predictive

Parts are castings and invariantly change in size. Parting line flashing varies +/- 0.1875 Example 4. Robotic Grinding. Steel castings presented to the robot. The excess flashing is required to be removed around the casting surfaces as the robot takes the part to the media (wheel)

Inputs (tolerances) I output (tolerances)

Parting line on the casting to be within +/-.030" of the base material

Solution to manage the tolerance variance between inputs and outputs: Fixtured inbound conveyor utilized to present queued raw material within tolerance for accurate part location Laser sensor attached to grinding wheel to offset robot path into wheel as the wheel wears in real time Grinding wheel attached to a compliant back stand that enables the robot to maintain a constant force when at the robot pick position the casting is presented to the wheel surface regardless of dimensional part tolerance or amount of flashing on the casting surface

FIG. 8 Stack up of Tolerances

======

It takes time to manage variability, and at a minimum, money.

There are many tools available to use in managing the stack-up of tolerances, but there use will incur additional cost and training. The thing to remember about validation is, that the final product constraints are constant -throughput, and product quality. The trick is to validate what is needed in regard to tolerances, and then validate whether the tools you decide to use can achieve the project expectations. Four examples described in Figs 8.1 through 8.4, illustrate this point for several applications. All the project examples are still working well in production today. The point of these examples is to illustrate the considerations that are present for any application in terms of process tolerances. At the start of each of the projects, the team had to analyze what were the variables and what were the final expectations. The team then had to manage the solution that enabled the process to work with zero defects. Costs for managing the variables while achieving the specified throughput were just part of the overall robotic system investment.

Managing the stacking up of tolerances is one of the critical components that require validation because a machine (robot) with zero senses (sight, touch, smell, hear, taste) is being forced to work with something variable, in producing something constant. Human beings make this task look easy, which is why understanding the challenges of using machines is so difficult. In our imperfect world, humans make decisions in microseconds that overcome the slightest of variances to achieve consistent end results. Robot technology has been adapted continually to match the decision-making skills that humans take for granted. The irony is that, with all the sensory capability of human operators, it’s impossible to achieve zero defects in a manually- operated process, whereas zero defects should be the benchmark with robotic automation.

Ultimately, a good or bad project should also be judged relative to the amount of investment and complexity in relation to the level of robotic decision-making and system design used to overcome poor process tolerances. On one hand, keeping the system simple is smart. On the other hand, using innovative tools that successfully manage tolerance mismatch, offers the user much more flexibility.

If you have the right shop floor personnel, with the right attitude toward supporting the robot system, you should invest in as much technology as possible to gain the maximum from the robot system.

====

Risks: Tack size varies and seems generally too Gap condition varies Overall weld joint tolerance as a function of the forming process, distortion during welding, and fit-up will exceed the %the weld wire diameter fit-up rule large The weld joint is in itself a tricky configuration Expectations of the finished product: 3/16" weld size throughout the perimeter Minimal weld spatter No visual weld defects Minimum post weld grinding Minimum distortion

FIG. 9 Example- Proof of Concept

====

Proof of Concept Case Studies

Proof of concept, sometimes called an engineering study, or feasibility study, serves a valuable purpose in proving or disproving whether a solution will work without designing and building an entire system. Proof of concept is used when no other form of validation tool can determine whether the solution will work. Even if vision, for example, has been used successfully, hundreds of times, if it has not been tested in specific relation to your work-piece, then how can you be comfortable with investment in the projected robot system without testing the product? Proof of concept is commonly used in the following scenarios: Where the robot process controls the properties of the finished product, such as in welding or material removal Where the set of unknowns and risks cannot be validated through conventional offline PC-based simulation tools Applying new technology where there isn't enough experience to validate the success rate in relation to the specific application robot vision or other, will work for the specific application Determining whether a tool, even a robot option such as For instance, a firm wants to determine whether its current line of architectural stainless steel assemblies, shown in FIG. 9, can be welded by means of robotics. The challenge is to achieve consistent weld size, weld profile, and a throughput of seven assemblies per hour. The requirement is for a 3/16-inch corner weld around the perimeter of the assembly. Notice the variability of fit-up and gap condition, and remember that the rule of thumb for robotic welding is that the weld joint tolerance needs to be half the width of the welding wire diameter. If this rule cannot be met, then additional tools in the form of robot options ( touch sensing, through arc-seam tracking and even the possibility of vision or laser guidance for joint location and tracking ) are generally required.

The robot is welding the perimeter of the assembly, which is a cover plate located on top of a container. Notice the size of the tacks and that, without testing, the cosmetic appearance of the robot weldment has to factor in the tack size. The proof of concept exercise will determine that the tack size needs to be reduced to achieve a good robotic weldment. The throughput is not as much of a concern because the weld calculator shows that throughput can be achieved based on the proposed welding travel speed. The judgment criteria for a go or no-go with this program is the resulting bead appearance and weld size, as well as the presence of weld spatter.

These three characteristics significantly affect the next step in the value stream, which is grinding and polishing. The proof of concept will validate whether the robot process will achieve the production expectations, given the set of variable inputs in terms of the fit-up, tack size, and gap condition. The expectations of the weld size, weld appearance, and weld spatter, define the production constraints.

Usually the proof of concept will consist of the robot, a tool (gripper), and other tools that enable the process to be performed (i.e. the process equipment, vision system, robot software options, and process consumables). The proof of concept takes time and resources, and as a result, robotic system integrators or research institutions are set up to manage the effort better than the user.

System integrators will generally credit all or a major portion of the proof of concept, in the event that a system is ultimately purchased.

The value of the proof of concept is that a lot of information can be gathered about the potential success of the project for a minimal investment relative to the actual system cost. Validation is everything when you can get it early, versus finding out whether the process will work on the production floor or at a demonstration on an integrators floor. The important steps in developing a proof of concept are as follows: Identify unknowns and risk through the audit and discovery phase Determine how the criteria of unknowns can be validated (similar application, offline simulation, other research)

Through process of elimination, whittle the unknowns and risks down to the short list of criteria that needs to be validated through the proof of concept Define the production requirements and expectations for the proof of concept Match the conditions of what has to be validated within the real world environment Determine the minimum amount of process tools that must be designed, built, and integrated, to validate the unknown or risk criteria needed to minimize cost and time The following examples describe where it has been practical to pursue proof of concept:

The first proof of concept example was developed to validate whether the robot system was capable of picking laser-cut steel blanks from the nest, as shown in FIG. 10. The specific risks were as follows: Each of the 16-gage steel blanks is tabbed to the main sheet by a single tab, and this tab has to be broken by the robot in order for the blank to be removed.

The cycle time for each cycle consists of identifying the candidate to be picked up, breaking the tab, confirming the tab is broken, and palletizing the blank to a table close to the main sheet. Simulation will only validate the robot motion and not the time to break the tab from the main sheet There is a considerable difference amongst blank geometries that the robot and robot tool will need to handle. How can the gripper concept be optimized to handle all the geometries, with minimum impact on throughput and cost? Some blanks are mild steel and some stainless, so the material variance adds another twist to the robot gripper solution, requiring either magnets or vacuum to pick up parts The proof of concept does not encompass any other aspect of the system because the remaining functionality has been demonstrated to work for this application. The remnants (skeleton) of the main sheet will also be required to be robotically removed but that is not an immediate risk for the project. The testing criteria is straightforward and specific to validating the three aforementioned risks. The validation included the following tasks:

1. Obtain prints of all the blank geometries and study them; Overlay blank prints and categorize products by similar geometries and size Examine each category for holes and slots in the blank "as-cut" for the purpose of determining where vacuum could be a limitation, because there would be a limited blank surface area to pick on with vacuum grippers. It’s also necessary to identify the weight of the heaviest blank size The results of task one will develop the gripper concept which, in this project, will be one vacuum gripper and three magnetic grippers for all the part styles. The vacuum gripper will be specially designed to position the array of vacuum cups to achieve a universal vacuum pattern for all the blanks that are compatible with vacuum picking. Cups will be arranged in quadrants on the gripper, and the gripper will be designed to allow independent control of the various quadrants, depending on the ID of the blank being picked.

Three magnetic grippers were needed because of unique slot and hole patterns on the some of the blank part styles.

2. Build a mock vacuum gripper, not a production gripper for picking one of the higher volume blank styles, also a part having more complex geometry. Obtain the use of a six-axis robot with appropriate payload and reach for the application.

The payload analysis will take into consideration the center of gravity for the largest gripper and blank when handled together.

Manually break tabs for a period of time and watch the operator perform this task for a period of time. Identify some of the most difficult blank styles to be removed by the operator, and select one of those parts for the validation.

The value of step two is to validate the material-handling equipment (robot and end of arm tool) for one of the more complex blanks in terms of breaking the tab that joins the blank to the main sheet. Selecting one of the more complex shapes, and studying the manual sequence, will offer a lot of value in terms of the robot programming routine that will be developed.

====

The robots task is to receive blank location and orientation from the translator software that converts nesting positional data to robot coordinates. The remnants (skeleton) will also be removed by the robot. Blanks will be sorted and palletized for the next operation in the value stream which is forming and welding Parts are tabbed to prevent shifting during handling of the main sheet

FIG. 10 Case Study Example- Robotic Sorting of Tabbed Nested Blanks

====

There are some assumptions that will be made and not validated.

The assumptions are that the tab for the stainless or steel blanks will behave in a manner similar enough to not require a completely different test for each base material. The stainless steel blanks will be tested in this program because the more-complex shapes are made of stainless. Additionally, focusing on a complex blank shape is going to validate the removal for all the other shape styles, without having to set up additional tests or make additional grippers for other blank shapes.

The proof of concept will thus include the robot, gripper, a table to hold "as-cut" blanks, and a table to allow the robot to place the blanks in specific patterns and orientations. The key to the robot programming will be to use the six-axis robot and tool to peel the blank at about a 45 degree angle to the main sheet but at varying speeds throughout the movement, while utilizing a motor-driven grinding wheel to break the tab as the robot is peeling away the blank. Additional work will be required in the proof of concept to confirm that the tab is broken, prior to the robot fully peeling the blank away from the main sheet.

Another example of situations where proof of concept can be invaluable is simply determining the "optimum" solution, because there are so many ways to automate a manufacturing process. For the situation under review there is no question that the robotic system will work. However, minimizing cost, keeping the solution as simple as possible, and providing the right amount of flexibility, all need to be investigated. The value of the proof of concept in this scenario allows for the best solution to come to surface, prior to making a major investment and a potential direction change, which will cost a lot more money and time.

====

System monitoring.

changing parts r Conveyor, palletizing to trays, managing inspected r parts that fail inspection Qty. of grippers, interaction with machine work-holding.

tolerance for loading Peripherals Integration with robot, selecting the optimum equipment for automation Casting # 1 loading machine tool 6 minute cutting time 75 pounds maximum Six (6) diameters ranging from 10" to 28"

* Part is turned over between machine process Three (3) jaw chuck work-holding at machine

* Grip on part outer diameter both machines 1 Radial orientation required prior to

* Cutting on two Mori Seiki lathes 1 Cell layout and robot configuration

- Casting # 2

* Radial orientation required prior to loading machine tool 4minute cutting time Cutting on two Mori Seiki lathes 25 pounds maximum

* Three (3) diameters ranging from 5" to 12"

- Part is turned over between machine process

FIG. 11 Proof of Concept Example- Robotic Machine Tending

====

The part attribute box for two steel castings that need to be machined in FIG. 11, illustrates the project attributes. Outside of the part attribute box are shown the criteria that need to be optimized to match the user's priorities through the proof of concept exercise. Some of the important decisions that will be considered, as the result of the proof of concept exercise, are as follows: Radial orientation of the steel castings is required prior to loading the lathe. It’s necessary to decide on the best way to present the casting to the system to 1) qualify the part location for picking, 2) optimize raw part queue time to minimize operator time in feeding the system, and 3) find the radial orientation of the casting prior to robot loading of the CNC lathe. The ideas that will be floating around are as follows:

3D vision bin for picking up parts from the container

delivered by the casting supplier

Conveyor with 2D vision for establishing single-part orientation and location

Conveyor with fixtured nests (pockets) to capture and The castings have multiple diameters, and it’s necessary to decide on the best way to have the system adapt easily from one diameter to another. All the choices above still apply, with the addition of how to design the gripping concept to adapt to the various casting diameters A second part design with several diameters is also going to be machined in the system at the same time as the first casting design. The choices are still all the same, except that they are compounded because two completely different part designs will be machined at the same time and they have completely different geometries, gripping surfaces, and machining processes. Two lathes will be used for machining one part design and two lathes will be used to machine the second part design.

Inspection and part marking is required for every five parts, for each part size. The equipment to be used includes four lathes, plus inspection and marking equipment, as well as provision for parts to enter and exit the system.

The proof of concept for this example will answer the choices the user has in regard to how the system will manage the flexibility requirements in machining the two part styles concurrently. The proof of concept will define the team's compromises, expectations, and needs. Additionally, the proof of concept will determine the ideal system layout to: 1) achieve the part throughput for each casting style, 2) stay within floor space constraints, and 3) incur minimum investment. Ultimately the systems objective is to make a good part. The criteria outside the attribute box all affect the system's ability to make a good part. As the team works through the exercise in validating the optimum system criteria to automate machining of the two casting families, the scope of work is the end result.

The next section will review the final phase of robot implementation prior to designing and building the system, which is the scope of work. The better the team prepares a scope of work, through the practice of audit/discovery, identifying risks/unknowns, and proof of concept, the higher the probability that the robotic system will be successful. Without following through on these steps, the robot system will never meet or exceed expectations.

The proof of concept is a powerful exercise, whether it's a simple simulation of a standard welding system, validating the optimum robot riser height so that the robot reach is slightly better for the programmer to optimize the weld torch angle for a few weldments, or full-blown application development. The life of the robot operator is greater than fifteen years, so additional time to validate the risks and unknowns, as well as chisel away to find the optimum solution is the best investment, beyond the monetary aspect, that the team can make.