Capstone: designing a palletizing cell

This capstone brings the course together in one design: an end-of-line robotic palletizing cell for a food manufacturer. Palletizing is one of the most common robot applications in both Canada and the United States, and it touches nearly every topic in ROB-440: choosing a robot type, designing a vacuum end effector and checking its payload, cycle-time analysis against takt, pallet pattern design, PLC sequencing and the robot handshake, networking and data reporting, safeguarding pallet exchange, maintenance planning and the business case. We work through the design step by step, as an integrator would for a concept proposal, then list the deliverables a complete design package would contain. The numbers are realistic but illustrative; a real project would confirm them with the customer's data, vendor tools and simulations.

Step 1: Requirements

The customer packs breakfast cereal into corrugated cases at the end of two packaging lines. Each line delivers up to 6 cases a minute, so the cell must handle up to 12 cases a minute combined. Cases measure 400 mm long, 300 mm wide and 250 mm high and weigh 12 kg. Each line runs a different product, so the two lines must be palletized onto separate pallets. Pallets are the North American standard 48 by 40 inch size, 1,219 by 1,016 mm. The customer's warehouse limits loaded pallet height, including the pallet, to 1.8 m, and the pallet itself is 150 mm high. The plant runs three 8-hour shifts, five days a week. Forklift drivers remove full pallets and bring empty ones. Today, two workers per shift stack cases by hand, a task that has caused back injuries. The customer wants a payback of under three years.

Step 2: Pallet pattern

The pallet pattern determines how many cases fit per layer and per pallet, the stability of the load, and the robot's motions. Cases are usually interlocked: alternate layers are rotated or mirrored so that cases bridge the joints of the layer below, like bricks in a wall. A good pattern uses as much of the pallet area as possible without overhanging its edges, since overhang damages cases in transport.

# Search simple block patterns for cases per layer on a pallet (no overhang)
from itertools import product
pallet = (1219, 1016)          # mm
case = (400, 300)              # mm
best = []
for a_len, a_wid in (case, case[::-1]):        # orientation of the main block
    for n1 in range(0, pallet[0] // a_len + 1):
        rest = pallet[0] - n1 * a_len               # strip left along the long edge
        main = n1 * (pallet[1] // a_wid)
        side = (rest // a_wid) * (pallet[1] // a_len)   # rotated cases in the remaining strip
        best.append((main + side, n1, (a_len, a_wid)))
best.sort(reverse=True)
count, n1, orient = best[0]
print(f"best simple two-block pattern: {count} cases per layer (main block {n1} columns of {orient})")
print(f"area use: {count * case[0] * case[1] / (pallet[0] * pallet[1]):.0%}")

best simple two-block pattern: 9 cases per layer (main block 3 columns of (400, 300))
area use: 87%

Step 3: Robot and cell layout

Several robot types could do this job. A four-axis articulated palletizing robot is the classic choice: high payload, fast, and the tool only needs to stay horizontal and rotate. A collaborative robot on a column lift could suit a single, slower line but would struggle with 12 cases a minute at 12 kg each. A gantry would work but occupies more floor space and height. We choose a four-axis palletizer serving two build stations side by side, one per line, with conveyors bringing cases from each line to a pick position between them.

The robot must reach the farthest corner of each pallet at the lowest layer and at the top layer with the tool fitted. With the robot base centered between pallets and about 1.6 m from each pallet's far corner, and the top of the stack 1.8 m above the floor, a palletizer with around 3 m reach and a vertical stroke covering the floor-level first layer to the top of the stack plus the tool's height is appropriate. A simulation confirms reach and checks that the arm clears the conveyors and fence.

Step 4: Cycle time

The takt for one robot serving both lines is 5 s per case at the combined peak of 12 cases a minute. Picking two cases at a time halves the number of robot cycles if cases arrive in pairs. A sequence chart with the robot vendor's simulation gives a typical pick-and-place cycle, from pick position to a place point on the pallet and back, of about 6.5 s for a double pick at mid-stack height, longer for the far corners and shorter for near positions.

# Cases that accumulate on one line while its station waits for a forklift exchange
import numpy as np
rng = np.random.default_rng(440)
rate = 6 / 60                     # cases per second from one line
exchange_s = 60                   # time to swap pallets once the forklift arrives
for mean_delay in (60, 180, 300):
    wait = rng.exponential(mean_delay, 100_000) + exchange_s      # station blocked, s
    cases = rng.poisson(rate * wait)                              # arrivals while blocked
    p95, p99 = np.percentile(cases, [95, 99])
    print(f"mean forklift delay {mean_delay:3d} s: 95% of exchanges need <= {p95:3.0f} cases "
          f"({p95 * 0.4:4.1f} m), 99% need <= {p99:3.0f} cases ({p99 * 0.4:4.1f} m)")

mean forklift delay  60 s: 95% of exchanges need <=  25 cases (10.0 m), 99% need <=  36 cases (14.4 m)
mean forklift delay 180 s: 95% of exchanges need <=  61 cases (24.4 m), 99% need <=  91 cases (36.4 m)
mean forklift delay 300 s: 95% of exchanges need <=  97 cases (38.8 m), 99% need <= 147 cases (58.8 m)

The simulation shows a common palletizing issue: if a forklift is late to exchange a full pallet, cases accumulate on the conveyor. At 6 cases a minute, every extra minute of delay needs about 6 cases of buffer, roughly 2.4 m of conveyor for 400 mm cases, and because delays have a long tail, covering the occasional very late forklift with conveyor alone becomes impractical when drivers are busy. Many cells therefore add an automatic empty-pallet dispenser and a full-pallet outfeed conveyor holding two or three loads, so that the robot never waits for a forklift and drivers can collect pallets when convenient. With that arrangement, the light curtains move to the outfeed conveyor's exit, which simplifies the safeguarding discussed in Step 6.

Step 5: Controls and integration

A PLC coordinates the cell. It controls the infeed conveyors and their accumulation zones, reads photo-eyes that confirm cases at the pick position, tells the robot which line and which pattern position to serve through the handshake described in Lesson 5, monitors vacuum levels through IO-Link switches, tracks case counts per pallet, and signals forklift drivers through stack lights and the plant's warehouse system when a pallet is full. Recipes for each product store the case dimensions, pattern and layer count, so that a new product is added by entering data rather than reprogramming the robot, using the palletizing software that most robot vendors provide.

Each finished pallet receives a label with a serial shipping container code, which links it in the warehouse management system to the product, lot and date, supporting food traceability requirements. The cell publishes production counts, OEE losses and fault codes to the plant's data system, following the naming conventions of the plant's namespace as described in Lesson 6, and the cell's network sits in its own security zone behind the plant firewall.

Step 6: Safeguarding

The hardest safeguarding problem in palletizing is pallet exchange: forklifts must enter the pallet stations, but people must not walk in while the robot is working there. The classic solution is a light curtain across each pallet station opening with muting: additional sensors recognize the pattern of a pallet or forklift passing through, and the safety controller temporarily mutes the light curtain only for that pattern, while a person walking through triggers a stop. Muting must be designed carefully so that a person cannot walk in beside or on the pallet, and the muting sensors must be placed so that they are triggered in the right sequence only by the load. An alternative is to let the robot keep working at the other station while the exchange station's opening is open, using safety-rated zone limiting so that the robot cannot enter the open station, which keeps the robot productive during exchanges.

The risk assessment follows ISO 12100 and ANSI/A3 R15.06-2025. For the gate used by maintenance staff to enter the cell, the analysis in Lesson 8 gives a high required performance level, so the gate uses a coded interlock with guard locking wired to a safety PLC, and the robot's safety stop input achieves the required PL. The light curtains and muting arrangement follow the manufacturer's instructions and the safeguarding performance criteria of ANSI B11.19. Lockout/tagout procedures for the robot, conveyors and vacuum supply meet OSHA 29 CFR 1910.147.

Step 7: Maintenance and reliability

The maintenance plan lists preventive tasks from the robot manual, such as gearbox lubrication intervals, encoder battery replacement and cable inspection, plus cell-specific tasks: inspecting and replacing vacuum pads, cleaning vacuum filters, which clog with corrugated board dust, checking photo-eye alignment, and verifying light curtain and muting function with a test procedure at each shift start. Condition monitoring uses vacuum build-up time per zone and the robot's axis torque trends, as Lesson 10 described. Spare parts kept on site include vacuum pads and filters, a vacuum generator, photo-eyes, a muting sensor set, and a spare tool if the plant has no other way to keep running during a tool failure.

Step 8: The business case

An integrator's budgetary estimate for this cell might be around $520,000, including the palletizer, vacuum tool, conveyors and accumulation, pallet station sensors, safeguarding, controls, engineering, installation and training, plus a 10 percent contingency. The cell replaces two workers per shift on three shifts, five days a week. Using a fully loaded labor cost of $34 an hour and 250 working days, the labor saving is about $408,000 a year; subtracting a technician's time, energy, vacuum pads and service of perhaps $45,000 a year, and adding avoided injury costs of $15,000, the net saving is roughly $378,000 a year. Simple payback is about 1.4 years, well inside the customer's target, and the after-tax position improves further with the first-year write-offs described in Lesson 11, subject to current rules and professional advice.

Step 9: The design package

A complete concept and detailed design for this cell would include the following deliverables, which together demonstrate the skills of the whole course.

Working through a design like this shows how strongly the parts depend on each other. Picking two cases at a time changed the tool, the payload, the cycle time and the vacuum zones. Letting the robot keep working during pallet exchange changed the safeguarding, the safety functions and the throughput. The customer's staffing plan changed the payback more than any technical choice. Industrial robotics rewards engineers who can hold the whole system in view.