Industrial Robotics: Lean Manufacturing With Robotics for Low Volume, Small Batch Runs (part 1)

Home | Articles | Forum | Glossary | Books

AMAZON multi-meters discounts AMAZON oscilloscope discounts

Sections two through five were primarily concerned with the best practices in implementing robotic automation. It was important to start this guide with the guidelines, rules, and considerations involved in making informed decisions about implementing robotics into the manufacturing value stream. The last two sections of this text are about how robotic automation can be applied to maximize profit, which we should all agree at this point is the single most important goal of any firm.

The statistics hold true that one out of ten manufacturers plan and implement a robotic system for some task, whether it be welding, machine tending, palletizing, or whatever. A very popular obstacle that is often heard is, "we are a job shop and we don't make enough volume of any particular part", or "labor is inexpensive, and plentiful". Last, and probably most common is, "we have too much changeover and robots are for high volume". So there is a common theme here in the perception that setup and changeover is not compatible with robotics, and that for robotics to be beneficial to the firm requires "high" production volumes. The premise behind these obstacles to automating is one of the factors that motivated me to write this guide, and specifically this section.

This section will set out to compare and contrast the presumed reality regarding small-batch runs and robots. Most of the discussion will be in terms of small batch to single piece, continuous flow, which requires the maximum flexibility. This passage could be titled "robotics for high changeover and small batch sizes'' because it will validate the exact opposite premise to what is a commonly- perceived hurdle to automating in the first place. The material will describe how robotics can be more compatible with changeovers for small-batch and single-piece continuous flow than for plants in which the daily production parts never change.

Let's go back to some basics. Lean Manufacturing is a model for optimizing a manufacturing process. Lean manufacturing strives to enable a facility to be in a position to ship different quantities and different products on a daily basis with zero defects, and by the way, be globally competitive. In the author's opinion, the concepts of the job shop model and lean are not the same. Lean is a manufacturing practice that is implemented, regardless of whether a firm is a job shop or produces the same product everyday. Lean is all about eliminating the non-value-added tasks from the process. In a lean shop there are two types of products on the floor; raw material and finished parts ready to be shipped. There is no work in process, nothing happens without an order, and the manufacturing stream is agile in that it is designed to match the customer delivery requirements.

The job shop model describes a situation where the firm will build anything for anyone, and never really knows what is coming next.

The successful job shops that implement lean are those that practice the following precepts:

They implement cellular, continuous-flow manufacturing, with single machines or multiple, linked machines

They develop modular manufacturing work cells that allow for incremental capacity adjustment as a function of customer demand

They implement the manufacturing work flow in terms of customer-driven demand (TAKT) time, to match customer delivery requirements

They focus on programs that repeat in some sense of frequency versus a completely random forecast of production, which allows for predicted changeover/setups versus random changeover/setups

They implement tools that reduce and minimize manufacturing changeover/setups from one batch to the next


FIG. 1 Example of Cellular Continuous Flow Machining Systems

Continuous flow machining line capable of machining (3) style parts; Single work-cell with two machining operations designed for a batch of one ( single piece flow )


In FIG. 1, the single work-cell is designed around a lean job shop model where product shipped one day is different from what is shipped the next day. The work-holding fixture is a common fifty- taper tool, and the work-cell can be loaded with up to eighty (80) different part designs at any one time. The system will run unmanned for the entire time it takes to machine the (80) different pieces, which is more than twenty-four (24) hours. Every day the system can accommodate (80) new part designs, and because of the common work-holding arrangements for the parts to be machined, the system is capable of single-piece flow.

The other system example in FIG. 1 is a series of machining work-cells that are arranged in machining operation sequence and connected by a transfer conveyor. The system is designed to handle three different style products, but again the work-holding is similar for each. No changeover or setup is needed for the robotic automation. Each work-cell is designed around a TAKT time that matches the customer demand, which is one part every 81 seconds. Multiple product styles can be in the system at the same time. The product starts in the raw state as a die casting, and at the end of the line the finished part is completely marked, inspected, packaged, and ready to ship. The line can be modified to handle additional future capacity with little interruption because each cell is independent and modular. In other words, equipment can be easily moved, relocated, or added, to reconfigure a work-cell over a weekend if needed.

Processes where the product varies enough that changeover and setup are a factor will be reviewed later in this Section. All that being said, lean still applies.

Of course, no firm can react to its customer base with high labor cost per hour, inefficient manufacturing flow, high rework rate, poor asset utilization, poor planning with either too much or too little inventory, wrong equipment, and untrained operators. The desirable key ingredients listed below can't be achieved to the fullest potential without the interaction of trained human operators and robotics.

Hiring low-cost labor may work for many firms, but there is still the cost of training, turnover, and the challenge of achieving zero defects. There is always someone in the world with lower labor costs. Labor cost should be a primary focus, as well as productivity gains to increase machine utilization or "on-process" time and reduce or eliminate the labor wage gap as in the welding example, The intangible benefits of enhanced quality, flexibility, agility, safety, consistency, right-sized equipment, cellular flow, and adapting to high changeover, can also contribute to reducing the cost per finished work-piece.

A few examples where industrial robotics overlap the lean manufacturing model are as follows:

Cellular manufacturing--Robot cells and manual operations coexist on the line

Quick changeover--Utilize the robot to manage as much of the changeover as possible through the programmability of robotics and complementary quick-change tools

Flexible Operations--The robot is a fully-programmable device, ideally suited for something new every cycle

Quality at the source--Robotics programmed to measure, inspect, and act, based on real-time feedback of product metrics

Pull Production--The robot will only perform a task based on an order. The robot will wait until raw material shows up at the work-cell inbound location

Production planning--The robot easily serves as the conductor in a work-cell, managing the production schedule

Automate redundant tasks -- Robotics performing redundant tasks increase asset utilization or "on- process time" versus human operators

In summary, whether the firm automates with robots or not, lean manufacturing practices will reduce costs and increase productivity as the manufacturing process evolves. Thousands of small and medium-sized manufacturers provide testimonials about how profitable they have become applying the lean model to their manufacturing operations.

Man vs. Machine

What prevents a firm from running with 100 percent efficiency.

As an example, pretend an engineer watched a process run for a single shift of production for the purpose of recording every second that was engaged in making the product and not making the product. Examples of criteria that contribute to downtime for a manual process might include:

  • Lunch and breaks
  • Waiting for product to arrive
  • Changeover/setup time in transitioning from making one part style to making the next
  • Efficiency degradation due to inconsistent operator availability to perform the task every time (i.e. people do require bathroom breaks right ?)
  • Varying effort to perform a task (some cycles will be slower than others for the simple fact that a human being will not consistently achieve the TAKT time every cycle throughout a shift)
  • Maintenance of equipment
  • Reworking product that was not made right the first time
  • Varying cycle time as a result of inconsistent raw material (i.e. slowing down weld travel speed for an increased gap condition on a weld joint )
  • Inspection

Now, compare man versus machine in relation to the aforementioned usual suspects of downtime. Back in Section 1, Chart 1-3 described the overall manual process efficiency for various applications. Sadly, for process-intensive applications such as material removal and arc welding, the process efficiency can be as low as 30 percent or lower. The highest rated efficiency was for manual machine tending at anywhere between 50 and 75 percent. There will be more discussion of machine tending later, because efficiency is certainly a function of the type of machine and the machining process. Although robotic automation is available to work 98 percent of the time, robot efficiency is degraded by the following criteria:

  • Raw material unavailable - waiting for product to arrive
  • Nowhere to place finished product
  • Change over setup time is related to non-robotic equipment
  • Equipment maintenance

It is interesting to note the differences in the characteristics that degrade both manual and robotic process efficiency. Two of the four categories listed above for robotic efficiency can be eliminated by simply setting up an efficient material flow. A benefit of robotics is that they expose inefficient material flow, whereas manual processes do not, because shop floor operators will compensate for product variation without necessarily notifying anyone. In the big picture of production time, tolerance time management is typically a small percentage because the firm invests in tools that manage the incoming tolerance variation for the robot. Tolerance is defined again as producing a consistent, compliant-finished part, considering the variability that exists among the raw materials.

Discussion of equipment maintenance is not included here because it is external to the comparison of man versus machine, assuming that the same assets are used for either scenario. One critical category is changeover and setup time during the transition from one part style to another. The perception is that robotics and small-batch runs or low-volume production, do not mix. It is necessary to define the criteria that can be involved with changeover.

  • Essentially, a new recipe or process configuration is involved. The recipe affects the following process attributes:
  • Part work-holding, which can be weld fixturing, machine tool fixtures, chucks in a lathe, or a generic device that positions the product repeat ably and rigidly during the manufacturing process.
  • Process equipment setup, which can be a routine or a set of routines and subroutines, that enable a piece of equipment to perform a task for a particular part type. For instance, part markers, press brakes, gages, shears, machine tools, and welding equipment, all have settings and programs for particular part styles
  • How raw material is presented to the process. Material may come in bins, tubs, dunnage, manually built and staged, or stacked on a pallet. In general terms, how product is fed into the work-cell has a significant impact on changeover parameters. So, when a batch run is completed of one product type, how does the process receive the new batch of raw material for the next run?
  • How finished product is transferred out of the work-cell. For instance, parts may be stacked into a container in a specific arrangement or pattern different from the previous batch just run, or in palletizing there may be a different pallet pattern from one product to the next

Regardless of a manual or robotic operator, the changeover requirements for a process will be the similar because they are specific to the customer's needs for that product. Lean helps either the manual or machine process to evolve and adapt from one product run to another with as little wasted time as possible. The manual process certainly lends itself to easy adaptation because of the human senses. For instance, it is easy for an operator to pick product out of a bin to load a machine tool, and that simple effort covers the inbound changeover. In comparison, the effort involved to enable a robot to pick product from a bin like a human is challenging. Changeover is taken for granted for a manual operation, but for a blind machine, changeover can become a nightmare. Available tools that can be used to enable a robot to act exactly like a human operator when it comes to transitioning from one part style to another will be discussed later.

Regardless of changeover or small-batch runs, the first question in comparing man versus machine is whether enough production hours are available to keep the robot busy. For example, an 800-piece, annual production forecast probably is not enough production to justify a robot from a purely financial point of view. Other motives may be involved that will make sense for the firm to invest.

The maximum amount of available production in a full year is 8760 hours. Rarely have robotic systems been utilized 8760 hours, but nonetheless, there are 8760 available hours to perform the "process". It is hard to say whether there is an absolute minimum number of production hours that are necessary to justify a robotic system, because there are many organizations that require robotics and have completely different sets of metrics that they use for robotic valuation other than productivity and labor savings.

Cost analysis has been used to examine the minimum production hours that will produce a three-year ROI for a robot system.

Approximately 1400 annual hours seems to be an approximate crossover point for achieving the three-year ROI, or 25 percent annual return on investment. This 1400 hours is associated with a single shift of 8 to 10 hours per day. There are 2050 hours per year available on an 8-hour single-shift basis, and the loss of the 600 hours is attributed to changeover(s), and maintenance activity. What is staggering is that if the firm used a manual operator on an %hour, single-shift basis, the operator would only contribute approximately 800 to 1000 hours of production, which means a greater than 50 percent penalty for using an operator. Lunch, breaks, missed work, and general lollygagging, are realities of the manual shop floor operator. Out of an 8- to 10-hour day the process only runs for about 50 percent of the time available. Granted, on some days the throughput percentage was higher, but it was also lower on other days. In a multiple-shift operation, the throughput varies by the shift, with second and third shifts generally lagging behind the first shift.

An interesting analysis is how much do robots improve the effects of downtime, waste, and inefficiency, and what is that improvement worth to the firm? Another interesting question is what if robotics were so much more efficient than a manual operation that setup time and changeover became irrelevant? Furthermore, what if the cycle time to produce a single piece was not very important because, at the end of the day, more parts were always produced? The bottom line is that the firm has a set of orders to make parts on a daily basis whether on a single-shift or on a multiple-shift basis. The orders need to be completed to fulfill customer demand as well as meet the quality expectations. Anything less than meeting this basic set of expectations costs the company profit. Firms sometimes become fixated on the wrong things. A classic example is the cycle time per piece. Cycle time is part of the equation, but arc-on time, machine spindle utilization, quality, etc. are often more important.

Changeover for Small Batches

Changeover and setup are relative to the extent that parts and processes vary. Obviously, the more variance, the more setup required. The more automated the system setup to save time and efficiency, the more the cost for the robot investment. Balances between flexibility and investment are always considerations for every robot system regardless of the application. Literally anything can be done with a robot and the old saying "it's only money" holds true. There is also truth in the perception that robotics are not compatible with small batch runs, especially in the area where an alternative form of automation is a less-expensive choice or where the programming time and the work-piece fixturing aspect in terms of robot grippers, or other part fixtures, become too expensive for the volume of production. These constraints usually prevail when the part designs are significantly different in geometry. Additionally, automation expense is not practical for one-off parts that are set up for one run and never run again.

When looking at the variations of part shape it is sometimes just not worth automating that one extra part, which is why the 80/20 rule works so well when evaluating an entire product-mix for robotic automation. The Pareto 80/20 rule has always been a good rule of thumb, and staying away from extremes is another good idea. It never fails that 20 percent of product yields 80 percent of the profit, and/or volume. In other words, if a firm can earn a good ROI on automating 20 percent of the product, and keep the robot busy for the 1400-hour minimum, then this example could work well.

Otherwise, evaluation of the product mix should be taken one product at a time and an attempt made to eliminate those products that absolutely require a brand new configuration. As more and more product is found to automate, and the company tries to automate everything, at some point it will always see diminishing returns on robotic automation. It is wise to continue manual processing of the unique or really low-running products.

Changeovers are inevitable when small batches are run. Changeovers and setups relative to process equipment are necessary evils, regardless of whether a manual or robotic operator is used.

Apart from the process equipment, changeover for the robot system can be completely manual or completely automated in most setups.

For example, in press tending, the robot can be programmed to change the brake tooling as part of the procedure. However, in machine tending, the robot probably cannot change chucks on a lathe, although it can certainly change tombstone fixtures and pallets on a machining center. The point is that the available robotic tools have enabled changeover to be effected painlessly through the use of vision, and other tools. There is a cost for this flexibility, but the costs are significantly less than they used to be. For example, the days of purchasing tooling and hardware that was dedicated to a specific part number are long gone. Automatic changeover of the robot program in press tending or palletizing is certainly another way to exploit the advantages of machine versus man. In machine tending and welding, however, changeover of the process equipment, primarily the work-holding fixture that secures the workpiece, usually requires the human touch. The time associated with swapping out a weld fixture or changing chuck jaws on a machine, is a wash between man versus machine.

When the firm requires automated changeovers to eliminate the possibility of human error within the setup, numerous tools are available. For example, bar code readers, WID tags, sensors, and robotic vision, can be used to read part identification so that the system can be reconfigured automatically. Often the term "recipe" is used as the definition of the system parameters that are required to be modified to support the configuration when a new part style is introduced into the system. The "recipe" can be reconfigured manually or completely automatically, and is driven by the product characteristics from one style to another as they enter the system as a batch of a single product or as a batch of "x" amount of product.

For instance, these characteristics might include:

  • Physical properties: weight, shape, base material, dimensional tolerance, and overall size
  • Finished properties: the final characteristics of the work-piece or finished product. In palletizing this code would mean unit load configuration of pattern, product per layer, and any slip or tier sheets as an example. Another example is surface finish in material removal, or bead size and weld penetration for welding.
  • What is the application? Some applications are more forgiving than others, regardless of how different the product may be. Palletizing is a great example of the application of universal tools to a very broad range of product numbers

The greater the change from one batch of parts to another, relative to the two aforementioned part characteristics is what drives the extent of the recipe complexity and then ultimately the risk, the cost, and the downtime involved in the setup. Some applications by their nature require longer setup times than others. For instance, press tending starts with a flat metal blank that is formed into a three-dimensional shape. Regardless of its other characteristics, the blank starts flat so that it is easier for the robot to handle. On the other hand, in machine tending, one batch of parts may be cylindrical, with an accessible internal diameter to grip on for loading the machine tool. The next part is cube-shaped, with irregular surfaces and essentially as different as night and day in geometrical shape, compared with the previous part.

Case palletizing is another example where, if the robot system handles cases all day long the robot gripper can be designed to handle a myriad of case sizes and weights. There will then be zero changeover because the cases are of similar geometry and they can be picked from their top surface with an array of vacuum cups. On the other hand, if the palletizing application requires palletizing of cases with a top, followed by cases with a lid that could fall off, then a changeover is required if the system is required to handle both product types.

Products that are similar in terms of their characteristics or their dimensions, can often be accommodated by the system using what is called parametric processing/programming. The same core program and process can be applied to the entire product mix because the manner of handling and processing are compatible with the core program, which is easily modified or adjusted through the use of offsets. Offsets are defined as incremental adjustments in terms of distance or a numerical value, relative to a nominal position or parameter value. Each time a new product is presented to the system, adjustments are made by an operator or automatically, to input the new product data. A new process recipe is then generated that adjusts for the new product. When the piece is finished and the system is ready for the next product, the system waits for the next adjustment.


FIG. 2 Example of a Highly Flexible Robotic System


An example of a parametrically configured system is a robotic palletizing system where the various product types are similar in how they are handled and simply vary in size. Robot grippers, robot path programs, and inbound and outbound hardware, all exhibit zero changeover from one product to another. Offsets in terms of the product location as they are located on the unit load are easily adjusted from one product run to another. Parametric programming obviously works best for a family of parts that can be handled the same way, transferred into and out of the system by the same method, and are subject to the same process. The products are scalable to each other, or at least have enough in common in terms of how they are produced that a common work-cell configuration can support the range of products. Parametric programming is certainly at the low end of the spectrum relative to the impact of changeover on a system, but families of parts are wonderful to automate and are less expensive on the scale of investment and risk to automate with robotics.

The recipe for FIG. 2 is configured after the product enters one of two locations. For the continuous single-piece flow, or a batch of one part, a vision system controlled by the robot determines part location and identification from the unique punch pattern in each blank type. The system does require the blanks to be right side up as they are transferred into the work cell. Once the part identification is made, the recipe is generated. The recipe for the system in FIG. 2 includes:

  • The robot programs for each robot to process the blank from raw to finished
  • The end of arm tool ( gripper ) to be used for the incoming part
  • The tooling for the press brake which is exchanged by the robot
  • The press brake program containing the bending sequences for the product

If a blank cannot be read for whatever reason at the inbound station, the robot will place the blank on a location for an operator to examine later. If blanks come into the cell through the low volume conveyor, the robot will wait to receive the part identification from the operator that transferred the material into the system. Otherwise, on the continuous batch of one conveyor there is no operator intervention.

Suggestions for the Design of a Robotic Work-Cell

Design robot systems are like manual systems with a cellular approach, where only a few or literally a single operation is completed in the cell. If multiple operations are required in a manufacturing stream, a series of cells can be arranged for an assembly-line approach with a conveyor or some other means of transferring the product from one cell to the next. The alternative would be to group a set of robots or operators into one location, but changeover in those situations becomes very difficult to manage due to the fact that the robots are dependent on each other like a group of operators, each with one leg tied to the next. Robots should be arranged within independent work-cells because error recovery and changeover is simpler than when they are combined. An exception to this rule is when multiple robots are linked together through robot technology that enables the robots to be programmed as a single set of coordinated axes.

Work-cells should be kept simple and focused on a specific task. With multiple work-cells in an assembly line, each work-cell needs to be designed to achieve the TAKT time that will enable the line to match the customer demand, which is the throughput at a rated efficiency. FIG. 3 shows an assembly-line type of system for welding and fabrication of an agricultural component. Each cell is sized with the right equipment and designed to meet the TAKT time of 18 minutes. This 18-minute TAKT time was derived from the premise that the customer requires 2.6 parts per hour, or one component every 22.6 minutes. For this example the firm selected 80 percent efficiency for the overall process, due to the need for changeover of weld fixtures, touch sensing to find weld locations, and movement of the product from one cell to another. Each work-cell on the line was designed for an 18-minute TAKT time. Eight different style products were run on the line, with the average batch size of 20. The line also included manual stations where it did not make sense to automate some of the welding due to complexity in fixturing and torch access to the weldments.

FIG. 3 Assembly Line Robotic Welding and Fabrication System

Within each work-cell balancing the tasks that changeover associated with the tasks in a work-cell is not overly complicated. The keep-it-simple, cellular principle, allows expansion for incremental capacity increases, meaning placing more robots at a particular operation when needed and they can easily be inserted. Spreading the overall manufacturing stream over several work-cells that are performing only a few tasks, is a better choice than using only one or two cells to do everything. In the future, if the firm needed more capacity or changeover, addition of a few simple work-cells instead of a major hardware-intensive work-cell would be much easier to manage and less expensive. Other advantages include quicker delivery and a shorter time frame to make changes to simpler work-cells, lower investment, and less programming and configuring.

Manual labor should be used to perform the more-challenging tasks, where variable product mix makes it difficult to justify automation of certain operations of the process. Additionally, manual operators should be paid to use human dexterity to accomplish tasks versus paying for the robot to perform the same dexterous motions. These operators would also be the line operators monitoring the robot work-cells, and kitting raw materials. This strategy will yield higher productivity and efficiency with less labor content.

Many firms will always have an operator present, even for a completely-unmanned system, and their model it to have the operator perform some of the tasks that can be expensive to automate, so that the robot can tackle the redundant work or perform inspection tasks.

Like robotics, lean manufacturing forces companies to use right-sized equipment in cellular modules that are designed around the customer demands. Robotics equipment is easily moved, removed, or added to the system, making the system extremely agile. FIG. 4 shows a typical flexible robotic system designed as a cellular assembly line. The system was designed around the philosophy of a batch of one, but additional robot systems can be added or deleted from the line. The system in FIG. 4 is designed to palletize three products per robot cell, with a unique part number being sorted ahead of the line using a bar code reader, and then transferring the product to a specific conveyor lane.


FIG. 4 System Example Designed Around Incremental Capacity

Direction of Travel of product; Example of a gripper that picks cases and divider sheets


The gripper illustration shows how the product is picked up at each inbound conveyor lane and how divider sheets (tier sheets) also are handled along with the product.

The world is not perfect and in manufacturing the stack up of tolerances is something with which robotics constantly have to cope.

Because of their inherent design for adaptability and compliance, and usefulness as tools to produce quality outputs from varying sets of inputs, robots have always been expected to be flexible, regardless of the application. This statement is true whether the robotic system is processing one part style or multiple styles. Nothing is ever really the same from one part to the next, so how can the need to automate batches of one, five, ten, or fifty parts change the rules of how robots work. Robots are programmable, adaptable machines - period. The next section will review large batch runs and infrequent changeovers. If the same product runs all the time, the irony is that the obstacle to automate with robotics is reinforced because a more dedicated and simple manipulator may be a better solution.

Changeover is frequent requirement, and robotics complements the firm's ability to implement lean manufacturing in order to produce different products in varying quantities, each day of production.

Automating the machining process - Case Study

The compatibility of robotics for small batch runs is shown by another example in FIG. 5, which shows a set of components for a machining application being considered for conversion to robotics or to be continued with manual tending. Tending machines is a process that can present a great deal of challenge in terms of changeovers versus other applications. The example in FIG. 5 takes the changeover example to an extreme because of the varying geometry of the parts.

In this installation, the firm needs to produce small batches of each part style on a single-shift, ten-hour basis. There is one hour's worth of breaks and lunch in the ten-hour shift. The work-cell contains two machines, each of which is capable of machining each of the part styles. The process will be arranged for the ability to run any part at any time, based on varying requirements coordinated on a weekly basis. Going back to the criteria for changeover, if the process maintained the current manual output, the following tasks are involved with changeover:

  • Fixtures at each machine tool need to be changed out, and the firm employs quick-change tooling. The maximum time to change over a fixture manually is about 15 minutes per machine
  • The operator will retrieve empty bins and supply full bins of parts
  • The operator will run the first part of each batch and inspect various dimensional features of the parts to check for dimensional compliance with customer specifications
  • A new machining program is selected from the machine tool library

FIG. 5 Varying Part Styles for Machine Tending Changeover

In summary the changeover time from one part style to another is approximately 25 minutes with the assistance of a fork truck operator in bringing the bins of raw parts and an empty bin for finished parts. Table 6-6 describes the TAKT time for each part at a rated efficiency of 100 percent machining efficiency.

Table 6. TAKT time for four part styles in machine tending example

For this exercise it is assumed that there will be two changeovers per day per machine, also that the ten-hour single shift yields a maximum of 2500 hours available for annual production.

The summary of downtime related to breaks, lunch, and changeovers in the work-cell is as follows:

Task Time (min.) Total time (min.)

(4) changeover tasks

(4) inspection tasks

(2) breaks

(1) lunch break

(6) cutting tool changes

25 minutes/setup

5 minutes

15 minutes

30 minutes

10 minutes

100 minutes

20 minutes

30 minutes

30 minutes

60 minutes

Total time: 240 minutes

Throughout the year of running these four part styles in the manual work-cell, the operator misses workdays because of sickness, and a few scheduled conflicts. For the days missed, another operator is required to run production, but that requires additional training costs for the firm. Also, when a closer look is taken at the number of finished parts per hour, it is found that the operator averages, on a daily basis, only 65 percent efficiency. Why is this efficiency not higher? For machine tending it is a documented fact that manual efficiency will be somewhere between 45-70 percent. In other words, when the machine is ready to be serviced by the operator removing a finished part and loading the next raw part, the machine is idle, waiting on the operator. As a result, over the course of a day, only 65 percent of the available production time after breaks and setup's for tooling changeover and inspection is productive. What happened to the other 35 percent of the available production time after the setups and breaks? It is lost through the human inability to perform a task at the same required speed, over and over again at every cycle.

In summary then, the manual production process operates as follows:

Total time on a daily single shift -- 10 hours

Scheduled downtime -- 4 hours

Time in production -- 0.65 * (10-4 hours) = 3.90 hours

Out of a ten hour shift then, the firm realizes 4.55 hours of net productivity. The hours noted above, when the customer's annual demand was reviewed, required 1792 hours (179 days). It was assumed that with the 10-hour shift there were 2500 available annual production hours or 250 days. The 4.55 hours of productivity per day are calculated by 3.90/10) = 0.39 days. The total days of production required to meet annual customer demand is thus 179 required production days divided by 0.39 production days = 458 days. This situation obviously presents a problem because there are only 250 production days in a year. The firm has the choice of running production on more machines with more operators on a single shift basis, or picking up an additional shift, again requiring additional operators. Most firms would adapt to the situation by throwing more operators and production hours at the problem. Although certainly not a way to lower costs, that will at least allow customer demand to be met.

If a robot loading system is applied to the above example, the considerations that are challenging for the robot system are as follows:

  • How to grip each of the four varying geometries of parts, the four part styles have different geometry
  • How to present the parts accurately to the robot, given that How to manage part inspection
  • How to manage where finished parts go, because the finished parts must not be allowed to touch each other
  • How to changeover the machine tool programs without operator intervention
  • How to arrange access to a machine for adjusting tool offsets when the cutting tools wear, as well as how to changeover cutting tools without stopping both machines running. The safety aspect of the operator and robot sharing the same space is a problem here
  • How to arrange for the system to receive some level of operator intervention needed to facilitate changeovers, inspection, feeding the system, and removing finished parts because a single gripper will not work

It is also a documented fact that robotics will provide 98 percent uptime, or in other words be available 90 percent of the time. The scheduled downtime, relative to the process of work-holding, changeover of cutting tools, and presenting new part styles, will be the same, comparing the manual and robotic process. Inefficiency linked to the operator in the form of machine asset utilization, inconsistency (wasted time), breaks, and lunch are criteria that can be exploited as a weakness by the robot. Robots do not take breaks, do not require training, insurance, or benefits, and never call in sick. In summary, for the robotic process, the robot is capable of producing:

Total time on a daily single shift -- 10 hours

Scheduled downtime -- 3 hours

Total daily productive time -- 0.92 * (10-3 hours) = 6.44 hours

On a daily basis then, for the single shift operation, the robot will increase production by 6.44 hours - 3.90 hours = 2.54 hours per day. Additionally, the system can be configured with an inbound system that will allow for a one-hour raw part queue. In developing an inbound system that allows for a one hour supply of raw parts, the firm can also run the system for one additional hour into the second shift unmanned, which amounts to daily unmanned production, less the cost of power and utilities for the system. Thus, the total robotic daily productivity contribution could be 7.44 hours. The 1792 production hours of required customer demand could be accomplished in 240 days, providing additional capacity of 10 days or 100 hours. The efficiency of the robot is estimated at 92 percent including the 8 percent for tool changes.

The interesting consideration here is that two machines can meet the production needs without requiring capital to acquire additional machines. The robot system provides almost twice as much productivity as a manual process but also reduces cost by decreasing labor, decreasing waste and inefficiency, and decreasing the cost of managing employee turnover and random days of missed work. Other considerations in terms of intangible benefits are the peace of mind that goes with having a consistent operator for at least 15 years, confidence in achieving improved customer satisfaction through consistent quality, and reliable daily output. The right decision for a firm that wins any new business is to challenge the manufacturing group in terms of achieving the maximum utilization of the firm's current assets through quick changeover methodology and robotics.

Savings over the cost of new machines alone, versus getting the most out of current machine assets with robotic automation is ample justification.

What is also interesting is that use of a stop watch to time the manual operator and the robot would show that the manual operator will be faster in producing a finished part than the robot on a given cycle. Does it matter if, in this situation, the robot is slightly slower for a given cycle? The robot system still has the padded additional annual capacity of 180 hours.

The foregoing example sounds pretty good, though there are certainly other criteria to examine in terms of the justification of this program. The "fear" factor of the unknown should not be an influence, especially given the potential productivity gains. What if the average part run was less than 10 parts, and there were significantly more part styles, each with its own unique geometrical features? There should be a formula into which the user could enter data and derive a yes or no decision as to whether the firm should pull the trigger on robotic automation.

As discussed in previous Sections there is a balance between how much flexibility is desirable in terms of managing changeover versus risk and cost. In the previous example, with only four part styles, and the fact that the firm would never achieve the asset utilization manually as it would with running robotically, the robotic program would probably make a lot of sense. Other costs and challenges to manage where the manufacturer runs small batch runs are as follows:

  • A unique gripper possibly for each part style. The cost of storing the grippers, and of the grippers themselves becomes prohibitive
  • Programming the system for all the part styles but also the parts that may be introduced, literally on a weekly basis, or even greater frequency
  • Interactions between the robot gripper and the work-holding fixture for each part style, because each part may have unique features where the work-holding clamps the part during the machining process. The more parts and work-holding fixtures, the more complicated and difficult it becomes to manage the entire process
  • Managing how parts are transferred into and out of the work cell

A family of parts, meaning a series of parts like cylindrical pulleys or shafts, lends itself to being easily automated, regardless of the batch size being run. The characteristics to look for that lend the parts to be grouped together with a common automation strategy in managing the aforementioned challenges are things like:

  • Common methods and surfaces for gripping (examples include the inner or outer diameter, or on the body itself if the part is cylindrical or shaft-shaped)
  • Common designs for machine (CNC) work-holding fixtures, chucks and collets
  • All the parts fit within a specific weight range and overall size. An example could be weights within 50 Ib. fitting into a cubic space of 24 x 24 x 24 inches

Families of parts lend themselves to be automated if the parts styles start with the same casting part number but have different features as a finished machined casting, or if casting features vary from a basic form. The robot changeover concerns the physical aspects of handling the product as soon it arrives at the system. The robot is a blind device without a human hand, and the robot is mimicking a human task which is the basis of managing changeover. Changeover is a reality though, and managing changeover cost effectively is a manufacturing competitive advantage. Perhaps, in the future, there will be a programmable device that resembles a human hand that can be integrated to the end of a robot arm and act like a real human hand instead of a more-dedicated gripper design.

Automation that can be incorporated into the machining cycle itself by the inherent design of the machine tool already exists. For example, a lathe already has the ability to gage parts and send itself tool offsets, mark parts, and de-burr parts, as well as check for tool life or broken tools. Additionally, if the user is machining a shaft- type part the lathe can be retrofitted with a bar feeder that essentially feeds "bulk" lengths of round "rough" bar stock into the lathe automatically. It is difficult to automate this machining process because the bar feeder attached to the lathe is already feeding "parts" to the machine. Usually, finished parts will fall down a chute out of the machine.

Unless some post-machining task needs to be applied to the finished part, the robotics value is reduced by the worth of the bar feeder and parts catcher. Tasks that have to be accomplished after machining always add to the justification of a robot because of the possibility of exploiting slower and inconsistent human endeavors.

For instance, inspection, deburring, washing, and marking are all post processes that can be easily automated with the initial investment of the robotic automation, the original purpose of which was to load and unload the machine. Many designs for robotics to service lathes include ways to allow the operator to access the front of the machine to change or adjust cutting tools and make minor changes to the machining program through the machine control system at the front of the lathe. Loading the lathe without having to open the door to get to the machine work-holding increases productivity because the door never has to open and close during the load/unload cycle.

FIG. 6 Various Examples of Robots Tending Machine Tool Models

Machining centers typically have rotary tables (pallet changers) or linear shuttles that allow the operator to unload/load finished and raw parts outside the machine while the machine is cutting metal. In this situation, the unload/load time is embedded within the machining cycle, increasing productivity and lowering costs. Machining centers with the capability of operating with the unload-load cycle internal to the machining cycle are good candidates for robotic automation. FIG. 6 shows examples of robots servicing various styles of machine tools.

cont. to part 2 >>

Top of Page

PREV   Next | Article Index | HOME