Designing a Robotic Assembly Line: A Technical Guide for Plants in Indonesia

Real payload: part, gripper, and distance to the centre of gravity

The payload figure in a catalogue is the maximum weight allowed to hang from the robot flange. What gets forgotten is that this weight has to cover the gripper, the part, and everything else attached at the end of the arm — hoses, cables, sensors, and the adapter plate.

An 8 kg part on a 12 kg gripper means the robot handles 20 kg. If the chosen robot is rated at 25 kg, that still fits on paper. But a second factor matters just as much: how far the load's centre of gravity sits from the flange.

A load reaching far out produces a far greater moment than the same load sitting close in. The robot may refuse to run at full speed, or run with a vibration that ruins repeatability. So what goes into the calculation is not only weight but the centre of gravity and the moment of inertia.

Our practice is to work out all three while the gripper is still a drawing, not after it has been built. A gripper 2 kg lighter is often cheaper than stepping up to the next robot class.

• The heaviest part to be handled, not the average
• Gripper weight complete with hoses, cables, sensors, and adapter plate
• Distance from the flange to the load's centre of gravity
• Moment of inertia, particularly for long or double grippers
• The acceleration needed to hit the cycle time, since dynamic load exceeds static load

Reach, working envelope, and when a seven-axis robot rescues the layout

Catalogue reach is measured from the robot centre to the furthest point with the arm fully extended. In that posture the robot is at its weakest and slowest, so a cell designed around maximum reach almost always disappoints.

What we look for is a cell where every working point sits in the robot's comfortable region: not so close to the body that the arm folds into an extreme posture, not so far out that it loses speed and stiffness.

The difficulty arrives when automation has to fit a layout nobody can change — a column in the way, a machine that cannot move, operator access that must be preserved. This is where a seven-axis robot offers what a six-axis arm cannot: the extra axis lets the arm fold around the obstruction and still reach the same point at the same orientation. For handling and assembly duty we use the GP Series, whose seven-axis variant exists for exactly this case.

The other option is to move the robot rather than the arm: mounting it on a rail, or hanging it from above or from a wall. All three change the shape of the working envelope fundamentally, and often solve a reach problem more cheaply than stepping up to a larger robot.

Grippers and EOAT: what decides consistency

If one part of a cell decides success more than any other, it is the tool at the end of the arm. A robot repeats its motion very precisely, but that precision is worthless if the part is not held in the same position every time.

How to hold it follows from the part's character. Vacuum cups suit flat, clean, non-porous surfaces. Mechanical grippers with fingers suit rigid parts with a clear gripping face. Magnets suit ferromagnetic parts that can tolerate residual magnetism. Flexible, oily, or porous parts almost always need a trial before the design is locked.

Just as important is how the part is presented to the robot. A jig that forces the part to sit in one position is far cheaper and more reliable than adding a vision system to correct a position that is never fixed. Solving it at the jig is the habit we hold to before adding a sensor.

For a cell handling several part types, the next decision is a universal gripper or interchangeable ones. A universal gripper avoids change time but tends to be heavier and larger; a tool changer adds complexity and a maintenance point, but keeps every part held the best way for it.

• Vacuum for flat, non-porous surfaces, with margin for small leaks
• Mechanical grippers for rigid parts with a clear gripping face
• Magnets for ferromagnetic parts that tolerate residual magnetism
• A jig that fixes part position before vision is even considered
• Sensors confirming the part is genuinely held before the robot moves
• A universal gripper or a tool changer, depending on how often part types change

Cycle time: break it down move by move rather than guessing the total

A cycle time promised without a move-by-move breakdown is a guess. The defensible method is to write out the full sequence — pick, lift, traverse, position, release, return — and give each step a time from simulation rather than from feel.

What gets missed is the time that does not belong to the robot: vacuum reaching full suction, gripper fingers closing, sensors confirming, and the neighbouring machine opening its door. In a fast cell the sum of these small times can exceed the robot's own travel time.

If the target is missed, the order to work through is: shorten travel by rearranging positions, overlap motion with another process that can run in parallel, and only then consider a faster robot. Stepping up a robot class is the most expensive route and often has the smallest effect.

On a line with several stations, output is set by the slowest station, not the fastest. Balancing load between stations often yields more output than speeding up one robot.

Synchronising the robot with conveyors and neighbouring machines

A cell rarely stands alone. It takes parts from something and hands them to something else, and how those three talk decides how often the line stops.

On an indexing conveyor what is needed is confirmation the part has come to rest before the robot moves. On a continuously moving conveyor the robot needs encoder position data to follow the part. The second is more complex but removes a stop that would otherwise repeat every cycle.

For machine tending, what needs agreeing is who waits for whom: the robot waits for the machine door to open, or the machine waits for confirmation the robot has cleared its working space. That agreement has to be explicit in both directions, including what happens when either side alarms mid-cycle.

What we recommend early is a small buffer between stations. A buffer holding a few parts stops a brief interruption at one station from halting the whole line, and it usually costs far less than the downtime it prevents.

• Part position confirmation before the robot starts to move
• Conveyor tracking for parts that travel without stopping
• A two-way agreement between robot and machine on who waits for whom
• Defined behaviour when either side alarms mid-cycle
• Buffers between stations so a brief fault does not propagate down the line
• Cycle data logged to the PLC or production system for traceability

Vision: when it is genuinely needed

Vision systems are often requested at the start of a project because they sound like the most modern answer. In practice vision solves one specific problem: a part position that cannot be fixed mechanically.

If parts arrive in a tray at a fixed position, or can be forced to sit in a jig, vision adds no value — only cost, setup time, and one more thing to maintain and calibrate. Lighting that changes through the day becomes another variable to control.

Vision earns its place when parts arrive scattered on a conveyor, when orientation is random, or when shapes vary so much a single jig is impossible. It also suits quality checks that genuinely require visual judgement, which is a different job from establishing position.

The order we recommend: settle as much as possible with jigs and part presentation, and use vision for what remains genuinely unsolvable mechanically.

Sealing and dispensing: robot and supply have to be sized together

Sealing and dispensing ask something different of a cell than handling does. What is judged here is not whether the part arrived in place but the shape of the material left on its surface: even in width, unbroken, and not pooling in corners.

The result comes from two things at once. From the robot: steady travel speed and a constant nozzle distance to the surface. From the material: pressure and flow that do not drift during the process, and for viscous materials, controlled temperature, since viscosity moves with it.

The most common mistake is setting material flow once and considering it done. The robot slows at every corner, and a constant flow pools there. The fix is to make flow follow travel speed, so the material laid per centimetre of path stays the same.

Because the result depends on robot and supply together, we treat them as one scope. Our background in industrial fluid handling is why we do not split a sealing application into two separate responsibilities.

Handover: what your team needs to run it without us

A cell that works on handover day but cannot be maintained by your team becomes a problem within months. So what is handed over is not only a working cell but the ability to run it.

What your team needs in hand: as-built layout and electrical drawings, the signal list between robot, PLC, and neighbouring machines, backups of the robot and PLC programs, and procedures for what will certainly happen — restarting after an emergency stop, changing a gripper, and recalibrating after a collision.

Often just as decisive is the consumables list with expected service life. Small components such as vacuum cups, seals, and encoder batteries are the most frequent cause of stoppages and the easiest to prevent by stocking them yourself.

Finally, people. At least one or two people at the plant need to be able to edit programs and handle routine maintenance. Without that, every small change to a part waits on an outside schedule, and the flexibility that justified buying a robot is exactly what gets lost.

• As-built layout and electrical drawings
• The signal list between robot, PLC, and neighbouring machines
• Robot and PLC program backups with a documented way to restore them
• Procedures for restart, gripper change, and recalibration
• A consumables list with expected service life
• At least one or two people trained to edit programs