How Does a Spray-Painting Robot Decide Where to Apply More Coat — and Where Less?(1)

Traditional manual spraying relies on "feel" — a seasoned painter intuitively knows where to lay on an extra pass and where to glide through lightly. When it comes to spray robots, that "feel" becomes an "algorithm." So how do you make a machine as smart as the old master?

Behind the answer is a complete closed-loop decision chain: 3D visual perception → surface geometry analysis → film-thickness distribution modeling → real-time feedback and correction. This article breaks down how that "brain" actually works.

I. From "Uniform" to "Appropriate" — Why Differentiated Film Thickness Matters

Many people's first impression of a spray robot is that it "sprays more evenly than a human" — but this is only half right. The robot's real advantage is not "averaging" — it's on-demand allocation.

"Uniform film thickness" ≠ "Appropriate film thickness"

Different areas of the same workpiece can have dramatically different coating requirements:

• Simply put: spraying every zone the same thickness is laziness; spraying each zone exactly what it needs is intelligence.

II. Five Decision Criteria — How Does the System "Know"?

An intelligent spray system doesn't make decisions based on a single piece of information. It relies on multi-modal data fusion across five dimensions:

Core logic: CAD design is the "base map," vision provides "corrections," the database gives the "starting position," wear prediction adds a "safety margin," and functional zoning sets the "hard limits."

The five factors are not simply stacked — they are weighted and fused. Visual data carries the highest real-time weight (because a workpiece can shift or deform), the CAD film-thickness targets set the floor (anti-corrosion requirements cannot be undercut), and functional zoning is the absolute red line (assembly interference is worse than a thickness overrun).

III. 3D Vision — "Seeing" the Workpiece Surface Is Just the First Step

In the previous article "The Eyes of the Spray Robot", we broke down how the visual system is mounted and how it forms images. In the context of film-thickness decisions, the vision system must go further: not just "see" — it must understand the surface geometry.

Key Point: Vision Isn't a One-Shot Scan — It Continuously Reads

Large workpieces (such as wind turbine blades and heavy-equipment body panels) can undergo subtle deformation due to their own weight, fixture clamping, or temperature changes. The intelligent system's vision module continuously captures data throughout the entire spray process, updating the surface model in real time — what was "there" a second ago may not be in the same place anymore. This requires the system's scan–registration–path-correction loop to respond in under 50 ms.

IV. Film-Thickness Distribution Algorithm — The Core Decision Logic of the "Brain"

Once the surface is recognized, the system moves to the most critical step: assigning a target film thickness to every zone and planning the spray parameters needed to achieve it. Several layers of algorithms are stacked:

Layer 1: Surface Compensation — Normal Tracking

This is the physical foundation for coating uniformity. The rate of paint deposition on a flat surface differs from that on a curved one — on a curved surface, the same flow rate covers a larger normal area, causing the actual film to be thinner automatically. The algorithm calculates the required flow increase based on the local radius of curvature:

Q_actual = Q_baseline × (1 + K / R)where Q_baseline is the flat-surface reference flow, R is the radius of curvature (mm), and K is an empirical coefficient related to the paint's characteristics (typically 10–50).

Layer 2: Edge Strategy — Boundary "Extension and Retraction"

Workpiece edges have a natural "escape effect" — spray mist that flies past an edge no longer deposits, causing the film at the edge to be thin. The algorithm handles each boundary zone as follows:

Layer 3: Internal Cavity Compensation — Counter-Measures to the Faraday Effect

In powder coating, internal corners and cavities suffer from the electrostatic shielding effect, reducing deposition (see the article on powder electrostatic adsorption). In liquid spraying, internal cavities mainly face the "shadow effect" — robot arm posture is limited, and spray mist cannot reach in a straight line. The compensation strategies differ completely:

Unified formula:

T_target (zone i) = T_baseline + ΔT_surface + ΔT_edge + ΔT_cavity + ΔT_wear

Each zone's target film thickness = baseline thickness + the weighted sum of four compensation amounts.

This formula is the mathematical expression of "where to spray thick, where to spray thin." Every compensation term has a clear physical meaning — it is calculated from surface geometry data, not guesswork.

V. From Target Thickness to Machine Parameters — Five-Dimensional Inverse Solving

Once each zone's target film thickness is determined, the next step is to back-calculate the spray parameters. This is an "inverse problem" — given the desired result (thickness), solve for the process inputs.

Five-Dimensional Parameter Linkage Table

Film thickness ≈ deposition rate × dwell time. Both deposition rate and dwell time are determined by five core parameters:

The five parameters are not adjusted independently — they are coupled.

For example, increasing flow while simultaneously slowing the trajectory can double the thickening effect, but may cause sagging. Increasing voltage while pulling the gun farther away may cancel each other out with no net efficiency gain. The algorithm must use a process constraint model (anti-sag, anti-pinhole, anti-orange-peel) to bound the five-dimensional parameter adjustments, ensuring that thickening attempts don't backfire.

VI. Real-Time Closed Loop — Correcting Problems Discovered Mid-Spray

The film-thickness distribution algorithm above produces a pre-planned schedule — essentially the "battle plan." But during actual spraying, there are too many variables: paint viscosity drift, nozzle micro-wear, air pressure fluctuations, ambient temperature and humidity changes — all requiring real-time sensing and adjustment.

Three Sensing Layers in the Closed Loop

The ultimate closed-loop goal: keep wet film thickness deviation within ±5 μm (corresponding to ±3 μm dry film), automatically correcting the previous pass's deviation before the next coat is applied.