|Home | Articles | Forum | Glossary | Books|
The reality is that robots and job shops have a lot more in common than their differences, and incorporating lean manufacturing into the model enables the system to further evolve into something special. In contrast, no one argues the fact that if a firm runs high volumes of a single product style or at least a series of similar products, automating production with robotics should be carried out for all the obvious reasons.
High volume, low changeover, means production that is set up perhaps once per day, week or month to run a single part style. In welding or material removal, this mode could also equate to welding a very large assembly for earth-moving equipment or polishing an impeller blade where the "on-time" process is in terms of several hours. Another example is robotically laser cladding of low-alloy components for the energy industry, where literally hundreds of welding passes are applied to the component, one layer after another. Of course another example is the automotive or OEM business where the number of piece parts per hour is high and the same part style is processed at each cycle. These examples consist of thousands of parts per month and certainly annually.
High volume or high throughput per hour means something different depending on the industry. In assembly, and especially the food industry, rates are staggeringly fast. In packaging, and high speed assembly, the robot pick/place motion in each cycle can be measured at rates in terms of less than a second to a couple of seconds per cycle. In the last section was discussed the large palletizing robot that had travel speeds enabling the robot to accomplish 8- 12 cycles per minute. As a result, the average cycle time for the palletizing robot is a 6-second cycle. With the constraint of 6 seconds, the robot throughput is increased by handling as much product as possible from multiple cases to layers of cases at a time, to achieve the highest throughput in each 6-six second cycle. For high speed applications, the robot gripper and the robot payload need to allow for picking and placing multiple products per cycle versus a single piece. Robots designed for high speed applications are capable of 60 cycles per minute as compared to the 8 to 12 cycles per minute of ordinary robots.
For high speed applications every little detail counts towards optimizing the rate. Examples of these details are as follows:
Location of robot relative to incoming and outgoing materials
-- How the robot is mounted (inverted, on a wall, or upright)
-- Travel distance of the robot from the pick up location(s) to placing the product
Robot axis speeds, and the ability to accelerate and decelerate quickly - robot motion is critical to ensure no wasted time
Time for gripper to actually grasp and release product
How many pieces can be picked up per cycle while staying within robot payload limits
Part orientation requirements. In other words, is additional time required to change orientation of incoming parts relative to the next location they're to be placed
Picking or processing product while moving or static
Waiting on a peripheral device to confirm handshaking signals with the robot as with detecting presence of a part, or confirm that the clamp is open or closed
f parts are moving while being picked up, a means for tracking and identifying part location as the piece travels is necessary (linear or circular tracking of moving conveyor for example)
FIG. 1 illustrates examples of high speed robot applications that were designed for high speed continuous throughput. The examples show product moving at a high rate along a continuously- moving line where the line speed sets the TAKT time of the process.
The picture with the red light shining on the conveyor shows robot vision that enables the robot to pick up product in a random orientation while the product is moving on the conveyor. Vision is another tool used to optimize the throughput of the work-cell. For any system, and especially for high throughput processes, the limitations of a single robot need to be understood, and then considerations for multiple robots, or multiple redundant cells, in order to achieve TAKT time of the line. High throughput requires redundancy, which means that if a system fault occurs, there is a backup system able to continue production, even if the rate is lower, until all the systems are back on line.
Robotic Machine Tending for High Production
High throughput for other applications, such as machine tending, might mean that a part is produced every few seconds. Extremely- fast cutting times (a few seconds) are typically not accomplished on a traditional lathe or machining center but rather require custom machines with dial tables. Dial tables allow multiple operations to occur simultaneously instead of sequentially. Engineer who evaluate the sequential tasks of a machine tending application, find that robots typically take approximately 6-9 seconds to load/unload a CNC lathe, excluding the door open and close time. This cycle time includes the robot movement in performing a part swap with the machine work-holding fixture. The robot gripper is designed to hold a raw part in one gripper and remove the finished part held in the machine's fixture with the other gripper. The robot motion then enables the raw part to be inserted within the machine's work-holding mechanism.
Regardless of whether it is on a lathe or a machining center, the part swap is a classic way of robotically tending a machine tool. The load/unload time will be longer because of the clamping and unclamping time of the work-holding, the machine door open/close time, and waiting to confirm the door is fully open or closed. A solid minimum for using a robot to service two CNC lathes in order to insure the robot does not become a bottleneck to the process and the machines are not waiting on the robot is 30 seconds. The longer the machining cycle, the more machines the robot can service, or be used to perform post process tasks within the cutting time. On machining centers a flushing cycle is often required after machining, the pallet changer needs index time, and time 1s needed for part clamping/unclamping. Robots can perform a part swap in as little as 15 seconds on machining centers, but for larger machines the load-unload time may be considerably longer. Machines processing engine block and cylinder head castings can take up to a minute to swap out a finished part and load a raw one into the tombstone's fixture.
Taking the machine door out of the equation entirely is certainly a benefit. Sometimes, on machining centers, users choose to leave the door open, especially when there is a wall between the machining area with coolant, and the outside loading area. On lathes, the door can be bypassed by putting the robot in the machine, if there is room, or mounting the robot on the machine above the work-holding chuck as with a CNC lathe. The robot is mounted on the machine and is able to extend its arm through a shutter door located on the top of the machine to reach the chuck area while the front door remains closed. FIG. 2 illustrates this example.
High throughput in terms of machine tending can be applied to a number of machines and operations, some of which are shown in FIG. 3. For instance, the cycle time may be 90 seconds of cutting time per piece, but when the annual production volumes require a work-piece to be machined every 8 seconds then it's easy to start adding up the number of machines required to produce this volume of output from a system that has a 90-second cutting time. The high-volume system example in FIG. 3 has the following characteristics: There are six operations labeled Op 10 through Op 60
The machining operations are denoted as Op 30 and there are (3) identical systems, each with the same quantity of machines and robots
Each of the machining cells contains four (4) machines and one (1) robot on a track
If one cell is down or faulty, the other two are capable of producing parts and the capacity of the line drops by one third
The robot on a track is capable of keeping up with four machines without becoming a bottleneck. A fifth machine would be difficult to accommodate in the cell
Twelve machines are required to achieve the required throughput, and the robot can keep up with four (4) machines, so there are three Op 30 machining work-cells
The advantage of this system is that one operator can tend all the machines. The operator's tasks include keeping the inbound queue supplied with raw material and managing removal of finished parts, as well as supporting various inspection duties. The cost of the automation in a work-cell is spread across the quantity of machines in the work-cell. The more machines that can be serviced by a robot, in addition to performing post process tasks such as deburring, marking, and inspection, the more the ROI is improved. The line in the sand where automation makes a slam dunk argument to be utilized is when the automation cost per spindle hits the $50,000 mark, or lower. The $50,000 has a high probability, because a single robot is here being used to service multiple machines with compatible cutting times.
When planning installation of a high throughput work-cell involving machine tools and including post-processing tasks such as gauging or marking, it is best to first gather the information about the required overall throughput of the system with the factored efficiency. For example, consider a throughput expressed as thirty (30) parts per hour at 85 percent efficiency. Then examine the cutting times for each machining process. The slowest machining time in the process is the pace setter. In other words a work-cell could have four machines comprising two CNC lathes and two horizontal machining centers. The cutting time for the CNC lathes is 120 seconds for one piece and 180 seconds for the same piece at the machining centers. The work-cell can't produce any more than two parts every 180 seconds because that is the slowest cutting time or, in other words, the pace setter. Once the throughput of the cell is confirmed, the next question is, "when the robot leaves the pace setting machine, can it accomplish all the tasks necessary including post process tasks and move back to the machining center in 4 minutes". To answer this question it is necessary to track the tasks after the robot leaves the pace setting machine and add up the time for each task until the robot returns to the pace setting machine. If the cumulative time allows the robot to return to the initial step within the 180-second period, then the robot is not a bottleneck in the process. FIG. 4 illustrates an example of the sequence of operations for a typical robotic machine-tending system.
The benchmark for the machine time is counted as door close to door open for a CNC lathe, or pallet index from side A to side B for a horizontal machining center. It is an important goal of robotic machine tending to ensure that the machine tool does not have to wait on the robot when its cycle is completed.
Part deburring, marking, gauging, inspection, and washing are just some value-added tasks that can be accomplished within the machining cycle. The investment in the machine tending is justified by the throughput gains and labor savings of the machine tending itself, and performing the value-added components is just gravy for the firm. Most engineers also think about how to utilize the robot beyond just the task of servicing the machine tool. Quality becomes a tangible competitive advantage when the thousands of parts that need to be made right the first time are considered. The cost of human error is amplified when the production volume increases.
Running Lights Out
Lights out means running the robotic system for a period of time without the intervention of a human operator to ensure safe operation of the system. Thus, no in-process inspection, or other manual decision-making can be introduced into the sequence of operations.
For example, there can be no part gauging within the machining cycle for checking part dimensions and adjusting program offsets throughout a shift. Another example would be automated weld inspection for the presence of welds and absence of visual defects.
An example of lights out production is when the process just takes a long time, such as in welding a very large assembly that takes several hours to weld robotically. Another example is an intensive material removal process such as in polishing an impeller blade for a wind farm. Because of the sheer nature of the "on-process" time, the robot system can run for hours without human intervention.
These types of programs are excellent candidates for automation because the amount of daily production time is so high that the firm can exploit the robot efficiency versus manual efficiency.
Additionally, major benefits are realized in terms of reduction in labor and more-consistent quality. The setup to present the product to the system is generally easy as well, relative to the amount of processing time. In other words, setup time versus how long the robot runs is a very small amount of time in the big picture of production. Most of the available production time is taken up by the robot performing the work rather than setting up parts or the processing system.
When the process cycle time is in terms of minutes, lights out working becomes more challenging. The challenge to lights out production with shorter cycle time processes is feeding the system and taking away the finished products. One of the goals when examining potential projects for robotic automation is how long it is practical to set up the system for lights out production. The time may be in terms of minutes, hours, or shifts. Decisions have to be made on what is affordable and practical in maximizing the time during which the operator requires intervention with the system. On average, many firms require 30 to 60 minutes of part queue before an operator has to reload the system with raw material. For large batch sizes, the inbound and outbound mechanisms that control material flow into and out of the system are critical because they affect how long the system can run without manual intervention, as well as a significant amount of cost of the system.
Some characteristics that affect how the material transfer is designed for unmanned production, or even a small raw part queue are as follows: Cycle time of the process. If the cycle time is in terms of seconds, then bulk feeding through bowl feeders or other types of feeding devices should be investigated. If cycle time is in terms of minutes then that changes things a bit and allows more flexibility such as the use of conveyors, or part pallets. On one hand you are dealing with hundreds of pieces and on the other with perhaps dozens.
Part shape, size, and weight. It is not practical to bowl feed some products, so other means such as a conveyor may be necessary to feed the system. The challenge is when the cycle time is in terms of seconds and the part can be practically bulk fed.
If the parts are heavy (exceeding 30 lb. and potentially into 100s of lb.) provisions are usually necessary to manage the size and weight involved. On the flip side, the process time is probably long, so fewer of the pieces need to be queued up Part Orientation requirements. Orientation requirements affect how the work-piece needs to be secured and located for accurate handling by the robot. Complex geometrical parts are a challenge to queue up due to the high costs of fixturing the parts so that the robot can "find" them.
How the parts come from the supplier; and what process is being automated. For example, in palletizing, it is a given that product is almost always conveyed from the upstream packaging area, and the palletizing is often the last sequence of a continuous moving line, so that the queue is built into the overall flow of material. In de-palletizing, the inbound material will be an entire pallet of product, and the robot peels away individual products or layers of product.
For arc welding, more firms are looking at using robots to handle the work-pieces and present them to other robots, but for more conventional welding automation, operators load and unload fixtures.
Without the operator loading and unloading fixtures, nothing happens. However, the disparity between robot and manual efficiency in welding is so significant that unmanned production does not need to be justified.
In machining, forming, and other machine-related processes, the chances are always increasing to set up work-cells for extended periods of unmanned (lights out) production. There are certainly the conventional means of conveyors, shuttles, and feeding systems, that require an operator for loading. The implementation of robotic vision has opened a whole new paradigm in staging raw material for extended hours of unmanned production for a broad set of applications
The use of vision for high volume production has decreased the investment cost and in fact simplified the implementation of robotic systems. FIG. 5 shows examples of how vision has enabled bulk quantities of parts to be staged without the high cost of handling every individual piece. The 'before' picture shows a network of conveyors with pallets, presenting the casting work-piece to the robot to provide the robot with the ability to pick up the piece repeatedly. The network of conveyors and the control system to manage it, certainly will work well in a lights out environment. The cons in implementing this style of part presentation for this example are cost, floor space, and maintenance. Also, new parts would require a new set of pallets. The 'after' picture is the same machining process, but with the conveyor network replaced by a simplified belt conveyor presenting loosely-located castings to the robot.
Robot vision provides the ease of determining part orientation and location without "dedicated" hardware to accomplish the same task. FIG. 6 implements vision in such a way that lights out manufacturing can be accomplished with little cost and setup. Bulk part presentation through the use of pallets, containers, bins, trays, and conveyor, all with loosely-located parts, allows a large amount of product to be queued up for hours of unmanned production.
Using robot vision for work-piece location can provide for considerable unmanned production with accompanying cost savings.
The divider board between the layers of parts shown in FIG. 6 is also picked up and set aside by the robot.
Ideally, when running large batches, a system can support an hour or longer of unmanned production. If orientation of the piece is required prior to processing, then vision should first be explored because of the ease of use and lower cost than that of dedicated hardware (i.e. "hardware" to capture the raw work-piece, such as transportation (i.e. the conveyor), and bowl feeders (with part escapements). Robots have the ability to handle dividers that separate layers of parts if they are palletized loosely on wooden pallets.
Robots can pick from bins, but there is a sacrifice of time while the robot determines the part location and the safest motion into and out of the bin.
The best manufacturing situations consist of many minutes of cutting time. Vision can be used to de-palletize parts from a bin or pick up simple parts that are loosely located on a pallet as shown in FIG. 6. The unmanned production is then extended into hours of machining time for little extra cost. The difficult situations consist of cycle times in terms of seconds, when part orientation is required. Running a cell unmanned for even 30 minutes can become complex and costly.
Job Shop vs. OEM
High-volume/low-changeover production for other applications such as welding or material removal certainly needs to be configured to achieve the throughput that best corresponds to customer demand, but with the minimum of fuss in terms of changing fixtures, robot grippers, etc. Where the drive for quick changeover was the last section's focus, here the focus is on producing parts in massive quantities, and as a result, time required for part transfer into and out of the work-cell is critical. The emphasis shifts from managing the system recipe so as to set up the system configuration for the next part style, to managing system uptime. Running large batches across multiple shifts requires a different approach in terms of insuring contingencies for downtime. The high-volume system is arranged to execute one procedure (recipe), but the slightest bit of downtime cannot be tolerated. Rarely now will a firm run high-volume production manually, precisely because quality and uptime are not to be achieved across multiple shifts consistently, using manual labor. The process has to be automated.
Below is a comparison of the priorities of the two modes of manufacturing :
High-change/low-volume (Job Shop)
Minimize setup time of peripherals
Make manual setup easy
Manage part recipes
System recipe management
[ Low-change/high-volume (OEM)
Maximize uptime of peripherals
Design for redundancy
Use of buffers
Keep process constantly moving
Ability to incrementally add capacity quickly
Maximize the raw material queue at the inbound
Discharge finished parts quickly as possible
Yes, there is overlap for low changeover or high changeover models in terms of the priorities, especially in maximizing the time the system can run unmanned without operator intervention feeding, and monitoring the system. However, in one situation the model is designed around the ability to make something new tomorrow, or even every cycle, versus designing for speed to produce one part style as quickly as possible. Another OEM criteria in common with the job shop model is that finished part inventory, and work in process, is a bad thing. These two items represent product that is not making profit for the firm. Both models require flexibility and adaptability in meeting the customer demands, whether it is changes in volume of parts or part styles. Neither model desires a lot of work in process, which drives both models to cellular manufacturing.
Manufacturing cells that are arranged within an organized flow that starts with raw materials from internal/external suppliers, and along the way each cell contributes a little value in producing the end product, is an outcome of lean manufacturing. Each work-cell is balanced relative to the piece per unit of time (i.e. pieces/hour) that matches customer demand. The key with a flexible cell using robotics is that the robot and system will not over-produce and is configured with the efficiency that insures high uptime and asset utilization. Additionally, if the work-cells are designed in a simple, modular configuration like a series of snap-together blocks, capacity can be increased or decreased by adding more blocks. The blocks can represent robot cells. The blocks can also represent components within a work-cell such as robots, and/or machines.
This concept is a key principle in the firm's ability to adapt to short-term fluctuations in customer demand and resulting production volume capacity needs. More equipment can be added when needed or equipment can be redeployed elsewhere, because the system is designed to be easily moved or altered when it contains a set of detachable building blocks. Some people would argue that robots are bad for situations where the capacity fluctuates a lot, because the firm paid for the robot equipment and there are times that the robot will not be workmg because there are no customer orders. However, it is necessary to look beyond the day that the robot is not working due to there being no demand or orders that the robot is configured to process, because the robot is an operator for 15 years plus.
Robotic Cellular Manufacturing
One more word about lean manufacturing and how robots complement lean principles for high- and low-volume production, and that is to develop work flow for a constantly-changing environment.
Work-flow should be configured so that it can be adapted easily to changes in capacity or product style. Series of work cells should be arranged with minimal and simple material transfer between the work-cells. Think of the work-cells as a series of operations along an assembly line. Allow for the line to accept robots, machines, or whatever peripherals are needed, quickly and effectively, and shun monolithic manufacturing processes or disjointed manufacturing flow.
Fabrication (laser, forming, bar, punch)
Manual weld booths
Work in process everywhere using containers to hold parts from on a process to another
In the machining area
Shears, press brakes manually tended
- Laser has material handling integrated within the laser process
Considerable in-balance from one department to another
Plant has dedicated inventory
Packaging and shipping area all manual
Painting/drying is accomplished using power free conveyor.
A lot of work
In process stored at every department in terms of capacity planning
Inventory is costing firm money everyday parts stay in inventory
Constantly between departments because
Painting is accomplished manually of geographic location to each other and different managers not knowing what each other is doing
Machining, deburring, washing, and part marking ail manual
Non compliant product is often not detected until assembly
Constant traffic jams and floor space problems with transporting raw and wip material throughout the manufacturing flow
FIG. 5a and b contrast bad versus good work flows and illustrate the point of arranging the manufacturing stream into a series of work-cells that pull product from one to the next, starting at the time an order is received.
The manufacturing flow in FIG. 5b allows for additional lines to be added to feed the main assembly and paint line or for additional cells within a particular commodity flow. There is no limit to how many factories are allowed within a factory or a commodity line that can feed the main assembly line. Cellular means a form of agility because cells can be added or subtracted and grown within the line to adapt to incremental capacity demands. Agile manufacturing is perhaps one more step beyond lean manufacturing because robotic "lights-out" processes are involved. Imagine the contrast between operators required to run the lean system utilizing multiple factory within a factory, or assembly line flow, versus designing for lights out production. A single operator could tend to an entire flow line in producing a commodity versus many operators required formerly. Additionally, the single operator can be performing on-the-fly inspection, resulting in higher quality at the source, which is another lean principle. The flexibility of this type of solution can include other forms of automation, or even manual processes for which robotics or automation are not justified, can be planted within the commodity flow line, working in tandem with the robotic cells.
All the work-cells, and the conveyor between the work-cells, are synchronized as a common fine-tuned machine producing to a balanced, customer-driven TAKT time. No inventory or work in process exists, other than the product in the machines themselves.
Ability to find problems before they travel way upstream is another major advantage of automation.
Many firms are set up exactly as depicted in FIG. 5b, and they are the global market leaders for their respective products.
Their cost of labor is less than 7 percent of the cost to produce the product, and they are all extremely profitable. Firms that incur less than 10 percent cost of labor in producing their products don't really have to worry very much about low cost labor. Manufacturing with robotics is a step above lean. Lean establishes the methodology and rules for batch-of-one production, or batches of thousands. Agile manufacturing enables a firm to become the low-cost producer period, against any labor rate in the world, especially when factoring quality into the justification equation. Robots also allow a firm to achieve zero defects.
Discussion of some of the tools used in manufacturing that enable the line to be planned for things that go wrong, break down, etc., certainly cannot be an afterthought. The first example is the principle of redundancy. Redundancy ensures that some amount of production can continue whenever a machine or a work-cell is down, for whatever reason. In other words, the firm duplicates the process in the form of an identical process, as in a duplicate work-cell or duplicate machinery within a cell. Basically, redundancy is design for contingencies due to downtime, to make sure that some amount of capacity is operational, based on unpredicted downtime.
Buffers are a way to allow the processes in a manufacturing line, whether it's continuous or pulse style, continue to build product in the event that an adjacent process is down. Imagine a manufacturing flow line containing ten steps. Raw material is introduced at step one. The material moves from step two to ten and eventually after step ten the product is complete. If step eight goes down for a reason, the last thing you want is for steps one through seven to stop. So the line will continue to build some level of product (buffer) that allows time to get step eight back on line. Or perhaps product is off-loaded from the system at step seven and is completed manually if the downtime at step eight exceeds the buffer time. Essentially, buffers buy the manufacturing line some limited amount of time that allows the rest of the system to process product in a normal fashion until the rest of the line can catch back up.
The Application between high-volume/low-changeover and low-volume/high-changeover
The implementation of a robotic system regardless of the high or low changeover criteria is the same. The basic principles that apply are:
By getting away from "fixed" components that are specifically designed around the characteristics of the work-piece, the user enables the work-cell to be inherently more flexible and accepting of future styles. The best advice is to design for flexibility wherever economically feasible.
The characteristics that enable the robot to be nicely compatible with high-changeover production are as follows:
We have now reviewed all the reasons why robotics is the ideal operator for a job shop practicing lean manufacturing. The reasons why robotics would not be ideal for high-changeover are as follows:
On the other hand, when there is a common, almost mundane, sense about the manufacturing stream, then the team has to examine all types of automation and not just robotics. There have been many projects for which robotics simply did not have the best mousetrap, because all that packaged flexibility in the six-axis robot, frankly was overkill. A "hard" or a more-dedicated form of automation makes more sense for a manufacturing process where everything is a constant. For example, welding round seams on a tank may not make sense for a robot system, and in fact, many firms believe that, and they implement a form of "hard automation," designed to turn the tank and direct the welding torch to the seam, and start the machine. Another example is in the packaging industry, where a continuous line of corrugated containers travels along and packages of "whatever" are inserted into the containers at a high rate. Literally in about every application from palletizing, material removal, assembly, machine tending there is a "hard" or dedicated form of automation that is offered as an alternative to a six-axis robot. Again, for some situations these devices are the right approach. Usually the firm's motivation is price related because dedicated or simpler automation tends to be less expensive. Hard automation devices tend to be less expensive because the firm is not paying for a six-axis manipulator such as the robot. Less axes of motion such as two or three axes may be sufficient for the application.
There are no wrong or right decisions about implementing either style of automation, and in many systems, both styles of automation are included. The implementation of robotics has certainly increased dramatically over the last few years at the expense of "hard" automation for all the aforementioned reasons. Additionally, instead of the robot industry experiencing inflation over the last twenty years, the opposite has happened, which obviously has been beneficial to the firm wanting to automate with robotics.
Sections 6 and 7 were written to illustrate how robotic automation could be useful to the firm running high-volume or low-volume production, with or without high changeovers. A good plan that identifies risk and benefits as the right way to start implementing a robotic system is reviewed in Sections 2 through 5. The planning process does not need to be complicated, but should be kept simple, at least for the first system. A project should be selected that will be a sure-fire winner, meaning that the confidence level is high in how the work-piece is produced from start to finish.
The success of the first robotic system will become a paradigm buster in terms of how the firm approaches manufacturing moving forward. In the closing sentence, this guide is certainly biased toward the implementation of industrial robotics as a primary contributor to improving US manufacturing. Not every program is compatible, but if the plant lights are on for about 8 hours a day making "something," then what does it cost to explore the possibilities of increasing productivity and lowering costs through the use of industrial robotic automation. Best of luck and thanks for innovating your business.