Industrial robots: simulate before you buy

Introduction : investir dans un robot industriel sans simuler, that was before!

In 2023, the global density reached 162 robots per 10,000 employees in the manufacturing industry, with major gaps depending on the country. France remains behind the most robotized countries, and this gap shows up in throughput, quality, and the ability to absorb variability. Source: International Federation of Robotics (IFR), World Robotics.

Key takeaway: the decision is not about one industrial robot; it is about the performance of an end-to-end flow, therefore about a validation through simulation.

I- Operational definition and scope: from an industrial robot to industrial robotization

Industrial robot: define the need clearly to decide based on measurable constraints

An industrial robot is a programmable manipulator that moves an end-effector in space to perform a repeatable task with a given level of accuracy. An industrial robot provides more stable cycle times, more consistent quality, and better risk exposure management—provided integration is done properly and data is reliable. The decision is made on measurable constraints: target throughput, product variability, tolerances, environment, and risk level. If one constraint stays vague, the project drifts in cost and schedule.

Industrial robotization: a system project, not the purchase of a standalone asset

Industrial robotization covers the industrial robot, the cell, machine safety, tooling, upstream and downstream flows, plus integration with information systems. A fast industrial robot does not compensate for unstable supply, a poorly sized buffer, or a non-standardized routing.

A robotized line is assessed by real throughput, OEE (overall equipment effectiveness), and scrap rate—not by nominal speed. The right decision scope includes commissioning time and ramp-up.

The must-have building blocks of a cell: axes, controller, end-effector, sensors, safety, upstream and downstream flows

The manipulator and its axes define reach, payload, and repeatability. The controller executes trajectories and manages I/O, with a direct impact on synchronization. The end-effector is used to grip, weld, screw, or paint; it fails when the scrap rate rises or when cycle time drifts. Sensors, vision, safety, and flows determine production robustness, therefore the ROI (return on investment) of an industrial robot.

Machine safety fails when risk assessment comes too late and forces layout rework. Upstream and downstream flows fail when real throughput locks onto the slowest link. Even the best industrial robot in the world cannot change a simple rule: a bottleneck sets the pace.

II- The “miracle machine” syndrome: why a catalog industrial robot generates unnecessary CAPEX

Nominal rate vs. real throughput: when 200 parts/hour becomes 150, then 120

The catalog shows a rate under ideal conditions, with implicit assumptions about feeding, gripping, part quality, and the absence of disturbances. In a factory, upstream rarely feeds at the theoretical pace, and downstream rarely absorbs without blockage. The outcome follows a common pattern: 200 parts/hour on the spec sheet becomes 150 with micro-stops, then 120 when buffers saturate. Profitability calculated on the nominal rate of an industrial robot is therefore a financial risk.

Bottlenecks, buffer stocks, and product variability: the three classic blind spots of datasheets

The first blind spot is the bottleneck, because a faster cell often moves the constraint instead of removing it.

The second is buffers, because too little buffer causes starvation, then too much buffer creates congestion.

The third is product variability, because a mix of references, tolerances, and surface conditions requires sensors, vision, or additional adjustments. Without flow modeling, an industrial robot datasheet tells a story, not a yield.

Hidden costs: tooling, integration, machine safety, data, maintenance, and training

The purchase price of an industrial robot rarely represents the total cost of an operational cell. Tooling, gripping, and commissioning quickly absorb weeks. Machine safety requires guards, scanners, interlocks, and validation through risk assessment, in line with applicable standards, for example the ISO 10218 series and ISO/TS 15066 via the European Agency for Safety and Health at Work (EU-OSHA). Data, maintenance, and training are underestimated: when these items are excluded from the business case, ROI degrades quietly.

III- What industrial robotics really changes on the shopfloor: OEE, quality, and cycle-time stability

OEE and automation: watch out for induced variability

OEE (overall equipment effectiveness) combines availability, performance, and quality to reflect useful output. Automation can increase variability if it creates longer stops during incidents, with more complex troubleshooting. Quality degrades when gripping or measurement remains unstable across a wide product mix. Useful robotization therefore requires discipline on data and process standards, including for an industrial robot.

Mini-case 1: machine tending, from manual to robotized

Situation: a machine-tending station handled loading and unloading of a machine tool: the robot (or the operator) picks a part upstream, positions it in the machine, starts the cycle, then retrieves the part at the exit. Production rhythm varied by operator.

Implementation: a robot cell standardized the pick, added presence detection, and synchronized the door opening with the PLC.

Result: cycle time improved by 18% in a stable configuration, and the stop rate linked to missed sequences dropped significantly. This held because raw material feeding was also upgraded; otherwise the industrial robot would have waited.

IV- Choosing the right type of industrial robot: six families, with “when to choose / when to avoid”

• Articulated robot: high versatility, footprint and commissioning constraints
  When to choose: if the task requires 3D trajectories, complex access, and product variants.
  When to avoid: if space is constrained or if commissioning must stay minimal.
  Order-of-magnitude benchmarks: payload from a few kilograms to several hundred; repeatability often around a tenth of a millimeter.
  The decision criterion is the real kinematics in the cell, not the theoretical reach of an industrial robot.

• SCARA robot: high speed and repeatability, limits on 3D and forces
  When to choose: if the operation is planar, fast, repetitive, with a high throughput requirement in a compact footprint.
  When to avoid: if the part requires high forces or complex 3D orientations. The decision criterion is read in the torque/force to be delivered and part position variability, therefore in the fit to the selected industrial robot.

• Delta robot: fast pick-and-place, payload and layout constraints
  When to choose: if the goal is very fast pick-and-place, with light parts and a stable flow.
  When to avoid: if payload increases or if the layout cannot provide the required height.
  The decision criterion is the pair “required rate + feeding stability”, rather than the datasheet of an industrial robot.

• Cartesian / gantry robot: working envelope and rigidity
  When to choose: if the need is a large work area, high stiffness, and a simple trajectory, often for palletizing or transfer.
  When to avoid: if floor space is limited or if format changes are frequent.
  The decision criterion is building footprint and ease of maintenance of the industrial robot.

• AMR and AGV: logistics flows and congestion risk
  When to choose: if internal logistics require regular feeding and transport traceability.
  When to avoid: if aisles are narrow or if traffic variability creates congestion.
  The decision criterion is traffic density and the robustness of priority rules, as much as the rest of industrial robotization.

• Multi-robot cells: throughput, synchronization, and maintenance complexity
  When to choose: if throughput requires parallelism or if several operations must run in cadence.
  When to avoid: if maintenance cannot keep up with complexity or if one robot stop stops all others. The decision criterion is the fallback architecture in case of failure and maintenance access, for each industrial robot.

V- Industrial robot or cobot (collaborative robot): decide with four decision variables

A cobot targets coexistence with humans, which often limits speed and allowable contact energy.

A non-collaborative industrial robot targets throughput, with physical guarding or appropriate detection devices.

The choice is decided on four variables:

1. target throughput and cycle time

2. frequency of human interaction

3. product variability

4. risk level

The trade-off is simple: higher productivity often requires more separation or tighter safety constraints.

Product variability drives an arbitration: a cobot can absorb changeovers if the human compensates, but that compensation costs pace. A faster industrial robot requires more consistent parts, or more sophisticated vision and tooling. Risk level requires a risk assessment and an architecture that reduces exposure, according to applicable standards. Safety is not handled at the end of the project, because it reshapes the cell, access, flows, and sometimes the throughput.

VI- Programming and operations: what changes after go-live

Three programming levels, three shopfloor realities

Point teaching works fast for simple motions, but it consumes hours of tuning as soon as references multiply. Offline programming builds and tests trajectories on a digital model; it succeeds if the geometric model reflects reality and if frames, tools, and tolerances are controlled. PLC (programmable logic controller) and MES (Manufacturing Execution System) integration changes daily life: standardized alarms, controlled stop/restart, clear loss reading. The decision criterion is the ability to connect industrial robot events to actionable causes.

Data checklist before the project: drawings, routings, tolerances, scrap, OEE baseline, and stop causes

The project requires up-to-date drawings, layouts, safety zones, and clearance volumes. It requires routings, cycle times per operation, tolerances, and explicit quality requirements. It requires scrap baselines, plus an OEE baseline with a stop-cause taxonomy. Finally, it requires a list of product variants and changeovers, to estimate the commissioning workload of an industrial robot.

VII- Simulation: the virtual crash-test of ROI before CAPEX

What simulation validates and how it arbitrates

Simulation tests mechanical, logical, and human interfaces between the cell and the rest of the line. It validates buffers and reveals dynamic saturation, because a resource close to its limit creates queues that grow fast. It turns an ROI hypothesis into a result tested against realistic scenarios: breakdowns, micro-stops, product mix, changeovers, and logistics disturbances, for each industrial robot and for the full flow.

An over-specified model that is underused costs a lot and produces little, which degrades ROI. A simpler, better-sized model can free global throughput if the bottleneck sits elsewhere or if integration reduces stops. Simulation enables this kind of arbitration, because it compares architectures on factory KPIs.

The decision criterion is total system throughput and total cost—not the datasheet of an industrial robot.

VIII- Virtual commissioning and digital twin: from ramp-up to supplier negotiation

Virtual commissioning: test the software before delivery

Virtual commissioning tests the control software on a digital model before equipment arrives. This approach limits on-site iterations, therefore production stops and urgent interventions. It can reduce on-site tuning time, sometimes cutting by half the ramp-up time when data is reliable and scope is clear. Without data discipline, commissioning shifts—it does not disappear, even with a high-performing industrial robot.

The digital twin as a supplier negotiation tool

A digital twin highlights blocking points: a congestion area, an insufficient buffer, or a fragile sequence. The discussion changes because the industrial player brings assumptions, scenarios, and metrics instead of absorbing a promise. Contracting relies on measurable targets—throughput, service level, scrap rate, changeover time—with test conditions and clear responsibilities between integrator and user.

The decision criterion is consistency between simulation, test protocol, and future operation of an industrial robot.

IX- Decision framework: the 5 traps to avoid before signing

• Trap 1: sizing based on supplier nominal rate, with no OEE baseline or measured variability
  Shopfloor symptom: the cell “runs fast” in demo, then waits for parts and multiplies stops in production.
  Countermeasure: build an OEE baseline, measure variability, then simulate throughput with breakdowns, micro-stops, and product mix before committing to an industrial robot.

• Trap 2: treating the end-effector as a detail, then discovering an unstable process
  Shopfloor symptom: random picks, marked parts, rising scrap, then permanent adjustments.
  Countermeasure: design the end-effector with the process, define tolerances, sensors, and material tests, then validate via simulation and prototypes.

• Trap 3: underestimating buffers, then moving the bottleneck to the wrong place

• Shopfloor symptom: alternation of starvation and congestion, WIP rising, then lead time drifting.
  Countermeasure: size buffers on variability scenarios, then verify dynamic saturation zones through simulation.

• Trap 4: delaying machine safety, then paying in delays and layout rework

• Shopfloor symptom: late addition of guards, impossible maintenance access, then commissioning delays.
  Countermeasure: start the risk assessment early and integrate safety into the layout, consistent with ISO 10218 and ISO/TS 15066.

• Trap 5: forgetting data and integration, then operating blind after start-up

• Shopfloor symptom: non-actionable alarms, unclear stop causes, no traceability, then invisible losses.
  Countermeasure: define events, tags, and PLC/MES integration before purchasing, to connect each OEE loss to an actionable cause on an industrial robot cell.

Dillygence uses the digital twin to test throughput, flows, machine safety, and ROI before CAPEX, with a fact-based approach that connects the industrial robot, the cell, and the factory system. Discover its Operation optimizer

FAQ: Succeeding with Industrial Robotization

1. Why shouldn't you rely only on “catalog speed”?

The catalog shows a rate under ideal conditions. In a factory, reality catches up quickly: micro-stops, variable part quality, and buffer saturation often drive throughput down. Going from 200 parts/hour on paper to 120 in real life is a common pattern if the flow is not modeled.

2. What is the difference between a robot and a “robot cell”?

The robot is the tool; the cell is the system. Successful robotization must integrate:

• The arm and its end-effector (gripper, torch, etc.) for the repeatable task.

• Machine safety (ISO 10218 standards) to protect people without blocking flow.

• Data integration (PLCs, MES) to understand why the robot stops.

3. How do you choose the right robot family? The choice depends on kinematics and the task:

• Articulated (6-axis): full versatility for complex trajectories.

• SCARA: ideal for ultra-fast planar assembly.

• Delta: the king of pick-and-place for light parts.

• Cobot: for direct coexistence with humans, at the expense of pure speed.

4. Why is simulation the ROI “crash-test”? Simulation tests mechanical and logical interfaces before CAPEX is defined. It reveals dynamic saturation: a fast robot can move a bottleneck instead of eliminating it. It turns a supplier promise into a result tested against realistic scenarios (breakdowns, mix, logistics disturbances).

5. What is “Virtual Commissioning”? It is virtual commissioning. Control software is tested on a digital model even before the robot is delivered. This can cut ramp-up time in half and avoid last-minute layout hacks on the shopfloor.

6. What are the three blind spots of robotics projects?

• The bottleneck: thinking a fast robot accelerates the factory (false if the bottleneck is elsewhere).

• Buffers: a poorly sized buffer starves the robot or creates congestion.

• Variability: a reference mix requires vision or complex tooling that impacts cycle time.

7. What are the 5 traps to avoid before signing?

• Sizing without a baseline: not knowing current OEE before robotizing.

• Neglecting the end-effector: the gripper/tooling often makes the process unstable.

• Delaying safety: late guards break the layout and maintenance access.

• Forgetting hidden costs: integration, programming, and expert maintenance.

• Operating blind: not planning stop-cause data collection from day one.