10. Process Monitoring, Control and Automation

Introduction and Context

Welding quality is determined not only by the selected process, consumables, joint design, and operator skill, but also by the stability with which process conditions are maintained. In manual welding, the operator continuously observes the arc, molten pool, joint geometry, and sound of the process, making rapid corrections. In mechanized and automated welding, these observations must be converted into measurable signals, interpreted by control algorithms, and linked to actuators that regulate welding conditions. Process monitoring and control therefore form the technical bridge between welding knowledge and repeatable production.

The industrial significance of this subject has increased with the use of high-strength steels, aluminium alloys, narrow-gap joints, additive manufacturing, large welded structures, and digitally integrated production systems. A weld may satisfy the nominal current and voltage specification yet contain defects because of poor fit-up, torch misalignment, changing joint gap, inadequate shielding, or thermal accumulation. Automation cannot eliminate these sources of variation unless the system is designed to detect and respond to them.

This chapter examines the architecture of welding monitoring and control systems, the signals generated during welding, signal-processing methods, adaptive control, mechanized equipment, industrial and collaborative robots, seam tracking, and offline programming. Emphasis is placed on engineering decisions: what should be measured, how measurements should be interpreted, which variables can be controlled, and how automation affects quality, productivity, safety, and organizational performance.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the distinction between monitoring, feedback control, adaptive control, and automation.

2. Identify important welding process variables and appropriate sensors for measuring them.

3. Describe signal conditioning, feature extraction, and decision-making in a monitoring system.

4. Select suitable control strategies for current, voltage, travel speed, wire feed, arc length, and heat input.

5. Explain the operation and application of mechanized welding systems, industrial robots, and collaborative robots.

6. Evaluate seam-tracking methods and recognize their limitations.

7. Develop a logical workflow for offline robot programming and process validation.

8. Analyse practical risks involving calibration, cybersecurity, safety, data quality, and human factors.

From Manual Skill to Closed-Loop Production

A welding system can be viewed as a chain consisting of a power source, wire or electrode delivery system, torch and tooling, workpiece, sensors, controller, and human or software supervision. The process produces measurable outputs such as arc current, arc voltage, sound, light emission, temperature, weld-pool geometry, and torch position. These outputs provide information about process stability and joint condition.

Monitoring is the measurement and interpretation of process information without necessarily changing the process. A monitoring system may record current and voltage and issue an alarm when values fall outside permitted limits. Control uses measured information to influence a process variable. For example, a controller may increase wire-feed speed when an increase in arc length causes voltage to rise. Feedback control compares a measured output with a desired reference and calculates corrective action. Adaptive control goes further by adjusting process parameters or control rules in response to changing conditions, such as joint gap, material thickness, or heat accumulation.

The distinction between open-loop and closed-loop operation is fundamental. In open-loop welding, parameters are established before production and remain unchanged during the weld. Such systems can be effective when joint preparation, material, and fixturing are highly consistent. In closed-loop welding, a sensor measures process behaviour and the controller modifies an actuator. Closed-loop operation is more capable of handling variation, but it introduces sensor noise, time delays, calibration requirements, software complexity, and additional failure modes.

A useful control-system representation contains five elements:

• a reference or set point, such as desired arc voltage or seam position;
• a sensor that measures the actual process state;
• signal conditioning and estimation that convert raw data into useful variables;
• a controller that determines corrective action; and
• an actuator, such as a travel drive, wire feeder, torch oscillator, or power source.

The difference between the desired and measured value is the error. A proportional–integral–derivative (PID) controller may use present error, accumulated error, and rate of change of error to calculate an actuator command. Welding systems often require filtering, limits, feed-forward compensation, and logic-based supervision in addition to PID control because the arc is nonlinear and changes rapidly.

Process Variables, Sensors, and Information Quality

The choice of sensor should begin with the decision that the system must make. A sensor is not valuable merely because it produces a large quantity of data. It is valuable when its signal is sufficiently reliable, timely, and physically related to a quality or control variable.

Electrical and Power-Supply Signals

Arc current and voltage are among the most widely monitored variables. Current is associated with melting rate and electromagnetic behaviour, while voltage is related to arc length and power-source operation. Their product, subject to appropriate interpretation, provides an indication of electrical power. Recording the mean, standard deviation, peak values, waveform frequency, and duration of excursions can reveal arc instability, wire stubbing, transfer-mode changes, or loss of shielding.

Electrical signals are relatively inexpensive to acquire, but they do not uniquely identify every defect. A voltage change may arise from torch movement, joint-gap variation, contact-tip condition, or a change in transfer mode. Consequently, electrical monitoring is strongest when combined with positional, thermal, visual, or acoustic information.

Position and Geometry Sensors

Encoders measure the position of robot axes, carriages, or wire-feed mechanisms. Linear displacement sensors and laser or structured-light systems can measure joint location, gap, groove shape, reinforcement, and bead geometry. Vision systems may use cameras with visible, infrared, or filtered optical wavelengths. Since the welding arc produces intense light and spatter, optical systems require suitable filters, shielding, exposure control, and robust image-processing algorithms.

A through-arc sensor estimates lateral joint position from changes in arc voltage or current while the torch oscillates. When the torch moves toward one side of a V-groove, the effective arc length and electrical response change. The controller uses the difference between signals on the two sides of the oscillation to calculate a correction. Through-arc tracking is economical and effective in many groove welds, but it depends on a detectable relationship between torch position and electrical response. It is less suitable where the groove is poorly defined or where process changes obscure the signal.

Thermal, Acoustic, and Optical Sensors

Infrared cameras, pyrometers, and thermocouples provide information about preheat, interpass temperature, cooling behaviour, and thermal distribution. Thermal monitoring is especially useful for materials sensitive to heat input or interpass temperature. Acoustic emission and arc-sound sensors can detect changes in metal transfer and process stability, although environmental noise and variations in shielding make interpretation difficult.

High-speed cameras and photodiodes can observe droplet transfer, arc shape, spatter, and molten-pool behaviour. These systems are valuable in research and advanced production, but their cost, data rate, optical contamination, and computational requirements must be considered.

Sensor Performance

Four characteristics are particularly important. Accuracy describes closeness to the true value; precision describes repeatability; resolution indicates the smallest detectable change; and response time determines whether the sensor is fast enough for control. A sensor may be accurate but too slow for arc-length regulation, or fast but too noisy for reliable defect classification. Calibration, electromagnetic compatibility, mechanical protection, and a defined maintenance schedule are essential parts of sensor deployment.

Signal Processing and Decision Architecture

Raw welding signals are rarely suitable for direct control. They may contain electrical interference, spikes from short circuits, mechanical vibration, optical saturation, or changes caused by legitimate process transitions. Signal processing converts raw measurements into stable and meaningful information.

A typical processing sequence includes sampling, isolation, filtering, normalization, feature extraction, and classification. Sampling frequency must be high enough to represent the relevant process dynamics. Low-pass filters may remove high-frequency noise, while moving averages reduce random variation. However, excessive filtering introduces delay and may conceal a rapidly developing fault. The designer must therefore balance noise reduction against responsiveness.

Features may include mean current, voltage variance, short-circuit frequency, arc-on time, thermal gradient, bead width, seam deviation, or the proportion of time that a signal remains outside a permitted band. Threshold logic is simple and transparent, but it may generate false alarms when process conditions naturally change. Statistical process control can identify drift by comparing measurements with control limits established from stable production. More advanced systems use pattern recognition, regression, neural networks, or other machine-learning methods to classify process states.

The decision architecture should distinguish between monitoring, warning, and shutdown. A minor voltage fluctuation may be recorded without intervention; repeated deviations may trigger a warning; a loss of shielding gas or severe wire collision may require immediate stopping. Alarm priorities, safe states, event recording, and operator acknowledgement should be specified before production begins.

A central concept is data traceability. The system should associate process data with a part, weld, program version, material batch, operator or robot cell, and inspection result. Traceability supports root-cause analysis and allows engineers to distinguish a genuine process change from a sensor or communication failure.

Control Strategies for Welding Processes

Control objectives differ according to the welding process. In gas metal arc welding, wire-feed speed, current, voltage, travel speed, inductance, and shielding-gas flow may be relevant. In gas tungsten arc welding, current, arc length, travel speed, filler addition, and shielding are important. In resistance welding, current, weld time, electrode force, displacement, and dynamic resistance may be monitored.

A common control loop regulates arc length indirectly through arc voltage. If voltage rises, the system may interpret this as an increase in arc length and move the torch closer to the workpiece. The control law must account for the response of the axis, measurement noise, and the possibility that voltage changed for another reason. A poorly tuned controller may oscillate, causing periodic changes in arc length and bead geometry.

Travel-speed control affects heat input per unit length, bead size, penetration, and production time. A constant-speed system is satisfactory where joint geometry is stable. Adaptive travel control may reduce speed when a larger gap is detected or increase speed where excess deposition is likely. Such changes must remain within qualified procedure limits; automatic correction is not permission to exceed the validated process window.

Wire-feed and current control are often coordinated with the power source. The controller may use synergic schedules in which a selected wire-feed speed automatically determines nominal voltage and waveform parameters. Adaptive systems can modify these values when feedback indicates a changing condition. Feed-forward control is useful when a disturbance is measurable before it affects the weld—for example, reducing speed in anticipation of a thicker section identified by a geometric sensor.

Control systems must include saturation limits, rate limits, fault detection, and recovery rules. If a controller reaches its maximum travel correction and the seam is still outside tolerance, it should not continue silently. It should record the condition and move to a defined safe response.

Mechanized Welding and Industrial Robots

Mechanized welding uses powered movement while an operator may still supervise setup, parameter selection, and process execution. Examples include tractor-mounted systems, column-and-boom machines, circumferential welding rigs, positioners, and dedicated gantries. Mechanization generally provides consistent travel speed and torch orientation, making it suitable for long, repetitive welds. Its performance depends strongly on fixturing, joint preparation, accessibility, and the ability to accommodate component variation.

Industrial robots provide multiple coordinated axes and can manipulate a torch around complex geometries. A robotic cell normally includes the robot, welding power source, wire feeder, torch-cleaning equipment, positioner, fixtures, sensors, safety devices, and a supervisory controller. The robot is not an isolated machine; its productivity depends on material presentation, loading and unloading, programming, maintenance, and inspection.

Robot selection should consider payload, reach, repeatability, axis configuration, torch accessibility, positioner synchronization, environmental protection, and service support. Repeatability does not guarantee weld accuracy. A robot may return precisely to a programmed point while the actual joint has moved because of fixture error, thermal distortion, or part variation. This is why calibration, probing, seam tracking, and process verification are required.

Collaborative robots, or cobots, are designed to operate with people under specified conditions using force, speed, separation, or monitored-stop functions. They can be useful for low-volume production, frequent product changeovers, and tasks such as torch positioning or handling. Welding remains a particularly demanding collaborative application because of arc radiation, heat, fumes, spatter, sharp components, and the need to control access. A cobot does not automatically make a welding cell safe. Risk assessment must address the complete system, including the torch, workpiece, positioner, fumes, electrical equipment, and possible unexpected motion.

Seam Tracking and Joint Location

Seam tracking compensates for a mismatch between the programmed weld path and the actual joint. The need arises from tolerances in cutting, forming, fit-up, fixturing, and thermal distortion. Tracking may be performed before welding, during welding, or through a combination of both.

Pre-weld laser scanning can map joint geometry and generate a corrected path. It is useful when the joint is visible and the geometry can be measured before arc initiation. Through-arc tracking operates during welding and avoids some optical problems, but it requires torch oscillation and an interpretable electrical response. Touch sensing uses electrical contact between the torch or a probing device and the workpiece to locate edges and surfaces; it is comparatively simple but slower and unsuitable for every geometry.

A robust seam-tracking system requires a coordinate transformation between sensor coordinates, torch coordinates, robot coordinates, and workpiece coordinates. Errors in tool-center-point calibration or sensor offset can produce systematic path errors. The system must also distinguish a real seam deviation from spatter, a temporary arc disturbance, or a missing component. Tracking gain, search amplitude, update rate, and correction limits should be validated experimentally.

Seam tracking corrects location; it does not automatically correct every weld-quality variable. A system may follow the centre of a joint while penetration remains inadequate because of incorrect current, travel speed, root opening, or shielding. This distinction is important in both system specification and acceptance testing.

Offline Programming and Virtual Commissioning

Offline programming (OLP) creates robot paths in a computer-based simulation environment rather than teaching every point directly on the shop floor. A digital model contains the robot, torch, positioner, fixtures, workpiece, safety envelope, and process constraints. The programmer defines approach points, weld paths, travel speeds, weaving patterns, process commands, and tool orientations.

An effective OLP workflow is iterative. First, the workpiece and cell are modelled using reliable geometric data. Next, the robot and external axes are selected and the tool centre point is defined. Paths are generated with appropriate lead-in, lead-out, orientation, and collision clearance. The program is then simulated for reachability, singularities, axis limits, cable interference, cycle time, and accessibility. Post-processing converts the virtual program into the controller language. Finally, the program is transferred under controlled procedures, verified at reduced speed, and validated with actual parts.

Digital simulation reduces programming time and production interruption, but it cannot remove the need for physical calibration. Differences between the digital model and the real cell arise from fixture tolerances, robot mastering, torch replacement, workpiece distortion, and coordinate-frame errors. Virtual commissioning is most effective when combined with accurate metrology, standardized tooling, and a disciplined change-control process.

Applied Case: Adaptive Robotic Welding of a Distorted Frame

Background

The following is an illustrative industrial case. A manufacturer produces welded structural frames from formed steel sections. The original robotic cell used fixed paths and constant travel speed. Although the nominal welding procedure was qualified, inspection showed inconsistent fillet size and occasional lack of fusion near frame corners.

Situation and Analysis

Investigation found that forming tolerances caused a variable root gap and that welding sequence produced distortion as the assembly progressed. The robot repeated its programmed path accurately, but the joint moved relative to the fixture. Electrical data showed occasional voltage excursions, yet voltage alone could not distinguish joint movement from transfer instability. The existing fixture also allowed small variations in part seating.

Approach

Engineers introduced three changes. A laser sensor measured joint location before welding and provided a corrected path. Through-arc tracking was added for selected welds where the joint remained visible to the electrical response. Current, voltage, wire-feed speed, travel speed, and seam correction were recorded for each frame. The controller was given correction limits and a stop condition when the required correction exceeded the qualified range.

The cell was recalibrated using a verified tool centre point and reference artefacts. A sample of frames was welded under controlled conditions to establish acceptable signal ranges. Operators were trained to interpret alarms rather than bypass them. The digital robot model was updated to reflect the revised fixture and sensor offsets.

Outcome

The combined system reduced path-placement variation and made abnormal assemblies visible before they progressed to final inspection. The most important improvement was not a single sensor but the integration of geometry measurement, process data, correction limits, and operator response. Frames outside the validated dimensional range were diverted for rework rather than being processed under uncontrolled corrections.

Learning Points

First, robot repeatability is different from weld accuracy. Second, monitoring data becomes useful only when it is connected to a decision and a defined response. Third, adaptive control should operate within a qualified process envelope. Finally, fixture quality and upstream dimensional control may provide greater benefit than adding increasingly complex algorithms.

Challenges, Risks, and Common Mistakes

Automation projects often fail because the process is automated before it is understood. If joint preparation is inconsistent, a robot may reproduce defects efficiently. Common technical mistakes include using uncalibrated sensors, placing sensors where they are contaminated by spatter, filtering signals so heavily that control becomes delayed, and treating a threshold alarm as proof of weld quality.

Another risk is excessive dependence on nominal digital models. The actual workpiece, fixture, torch, and robot must be measured and referenced. Tool-centre-point errors, incorrect coordinate frames, cable drag, and positioner backlash can all create path deviations.

Control systems may also become unstable when gains are too high, actuator response is slow, or multiple loops interact. Commissioning should therefore proceed gradually: verify sensors, test actuators independently, operate at low speed, introduce the control loop, and then evaluate disturbances.

Data-driven and machine-learning systems introduce additional risks. A model trained on a narrow range of materials or parameter settings may perform poorly when consumables, surface condition, joint geometry, or lighting changes. Training data should be representative, labels should be traceable to reliable inspection, and model performance should be monitored after deployment. Automated decisions affecting product acceptance should remain subject to engineering governance and defined validation.

Safety, Governance, and Professional Responsibility

Welding automation must comply with applicable machinery, electrical, robot, fume-control, pressure-equipment, and occupational-safety requirements. A risk assessment should consider normal production, teaching, maintenance, fault recovery, manual loading, unexpected restart, stored energy, arc radiation, fumes, noise, hot surfaces, and interaction with adjacent equipment.

Safety functions should be independent and testable where required. Interlocked doors, light curtains, emergency stops, safe speed functions, and protective enclosures must be maintained and periodically verified. Collaborative operation requires a documented assessment of contact forces, tooling hazards, workpiece geometry, and the effect of welding equipment on the collaborative risk profile.

Professional responsibility also includes preserving the integrity of process records. Data should not be altered to conceal alarms or deviations. Program changes, parameter changes, sensor replacement, and calibration results should be controlled and documented. Networked welding cells require access control, backups, secure remote access, and protection against unauthorized program modification. Cybersecurity is therefore a production-quality issue as well as an information-technology issue.

Current Developments and Best Practice

Contemporary welding cells increasingly combine digital twins, edge computing, high-speed sensing, cloud or plant-level analytics, and model-based quality prediction. Artificial intelligence is being investigated for bead-shape estimation, defect classification, parameter recommendation, and predictive maintenance. Its most responsible use is usually as decision support or supervised adaptation rather than unrestricted autonomous control.

Important developments include sensor fusion, in which electrical, optical, thermal, and positional signals are interpreted together; in-process three-dimensional measurement of bead geometry; and production systems that connect welding data with manufacturing execution and quality-management systems. These developments support closed-loop quality strategies, but they increase requirements for data standards, calibration, validation, and cybersecurity.

Best practice can be summarized as follows:

1. Define the quality characteristic and control decision before selecting a sensor.

2. Stabilize joint preparation, fixturing, consumables, and shielding before introducing advanced control.

3. Calibrate the complete measurement chain, including coordinate transformations.

4. Validate monitoring signals against independent inspection and metallurgical evidence.

5. Use correction limits, safe states, alarms, and clear operator responses.

6. Treat robot programs, sensor configurations, and models as controlled engineering assets.

7. Design for maintenance, cleaning, recalibration, and graceful failure.

8. Evaluate automation by total process performance, not robot utilization alone.

Chapter Summary

Process monitoring converts welding behaviour into measurable information, while control uses that information to regulate process variables. Closed-loop and adaptive systems can compensate for joint variation, thermal effects, and disturbances, but they require suitable sensors, signal processing, calibration, control limits, and validated procedures. Electrical, positional, optical, thermal, and acoustic signals each provide useful but incomplete evidence.

Mechanized systems improve consistency in repetitive welds, whereas industrial robots provide flexible multi-axis motion. Collaborative robots can support adaptable production but do not remove the need for comprehensive safety assessment. Seam tracking corrects path location, while offline programming reduces shop-floor programming time and enables simulation; neither eliminates physical calibration and process validation. Effective automation integrates equipment, data, people, quality systems, and governance. The central engineering principle is that automation should be applied within a known and controlled process envelope, with abnormal conditions made visible rather than hidden.

Key Terms

Adaptive control: Automatic adjustment of process parameters or control rules in response to changing process conditions.

Closed-loop control: Control in which measured output is compared with a reference and used to calculate corrective action.

Collaborative robot: A robot designed for defined forms of direct human interaction under specified safety conditions.

Control limit: A boundary defining the permitted range of a controlled variable or correction.

Data traceability: Ability to associate process data with a particular part, operation, program, and inspection result.

Feed-forward control: Control that compensates for a measured disturbance before it significantly affects the output.

Feedback control: Regulation based on the difference between desired and measured process states.

Offline programming: Creation and simulation of robot programs in a computer environment before physical execution.

Process monitoring: Measurement and interpretation of process conditions without necessarily modifying them.

Signal conditioning: Filtering, isolation, scaling, and preparation of raw sensor signals for analysis or control.

Seam tracking: Automatic correction of torch path to follow the actual joint location.

Sensor fusion: Combined interpretation of information from multiple sensor types.

Tool centre point: Defined reference point on a tool, such as the welding torch, used for robot motion calculations.

Through-arc tracking: Seam tracking based on electrical changes produced as the torch oscillates across a joint.

PID controller: Controller using proportional, integral, and derivative components of control error.

Virtual commissioning: Testing and verification of an automation system using a digital model before or alongside physical commissioning.

Review Questions

Short Answer Questions

1. Distinguish between monitoring, feedback control, and adaptive control.

2. Why are current and voltage signals useful but insufficient by themselves for complete weld-quality assessment?

3. Define tool-centre-point calibration and explain its importance in robotic welding.

4. What is the purpose of signal filtering, and what disadvantage may excessive filtering create?

5. State two limitations of through-arc seam tracking.

Long Answer Questions

1. Explain the architecture of an automated welding control system from sensing through actuation.

2. Discuss the selection, performance requirements, and limitations of electrical, optical, thermal, acoustic, and positional sensors.

3. Explain how offline programming is performed and why physical calibration remains necessary.

4. Compare mechanized welding, industrial robots, and collaborative robots in terms of flexibility, productivity, safety, and implementation requirements.

5. Describe how adaptive control can regulate travel speed, arc length, and wire-feed conditions while remaining within a qualified welding procedure.

Analytical / Application Questions

1. A robotic groove weld shows intermittent voltage increases and lack of fusion near regions with variable root gap. Propose a monitoring and control strategy, including sensors, signal features, corrective actions, and stop conditions.

2. A company plans to install a collaborative robot for welding in a shared workspace. Develop a risk-based implementation plan covering tooling, arc radiation, fumes, motion, human interaction, validation, and operator training.

3. A laser seam-tracking system gives inconsistent corrections because of spatter and changing surface reflectivity. Analyse possible causes and recommend a combined sensing, signal-processing, calibration, and maintenance solution.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes adaptive control from ordinary monitoring?

A. Adaptive control records process data without influencing the welding process.

B. Adaptive control changes process parameters in response to detected process conditions.

C. Adaptive control eliminates the need for calibration.

D. Adaptive control is used only for offline robot programming.

Correct Answer: B. Adaptive control changes process parameters in response to detected process conditions.

Question 2

A welding controller receives a voltage signal with significant random noise. Which action is generally appropriate before calculating a control error?

A. Remove all signal variation using an unlimited time delay.

B. Apply suitable signal conditioning while preserving the dynamics needed for control.

C. Replace the voltage signal with a fixed nominal value.

D. Increase controller gain until the noise disappears.

Correct Answer: B. Apply suitable signal conditioning while preserving the dynamics needed for control.

Question 3

Why does robot repeatability not guarantee weld-path accuracy?

A. Robots cannot execute the same program twice.

B. Repeatability concerns return to a programmed position, while fixture, calibration, and workpiece errors may shift the actual joint.

C. Robot repeatability applies only to mechanical handling, never welding.

D. Accuracy is determined exclusively by welding current.

Correct Answer: B. Repeatability concerns return to a programmed position, while fixture, calibration, and workpiece errors may shift the actual joint.

Question 4

Which sensor approach estimates joint position from electrical changes while the torch oscillates across the groove?

A. Touch sensing

B. Through-arc tracking

C. Infrared thermography

D. Acoustic emission monitoring

Correct Answer: B. Through-arc tracking

Question 5

A controller reaches its maximum permitted seam correction, but the joint remains outside tolerance. What is the most appropriate system response?

A. Continue increasing the correction without limit.

B. Silently complete the weld and delete the event record.

C. Enter a defined alarm or safe-state response and record the deviation.

D. Increase filtering until the deviation is no longer detected.

Correct Answer: C. Enter a defined alarm or safe-state response and record the deviation.

Question 6

What is a principal advantage of offline programming?

A. It makes physical calibration unnecessary.

B. It permits path development and collision analysis away from the production cell.

C. It guarantees that all workpieces have identical geometry.

D. It removes the need for process qualification.

Correct Answer: B. It permits path development and collision analysis away from the production cell.

Question 7

Which situation is most suitable for fixed-speed mechanized welding rather than complex adaptive control?

A. Highly variable joint gaps and severe distortion

B. Stable joint preparation, consistent fixturing, and a repetitive long weld

C. Unknown workpiece geometry and frequent product changes

D. A process with no qualified parameter range

Correct Answer: B. Stable joint preparation, consistent fixturing, and a repetitive long weld

Question 8

Why should a machine-learning monitoring model be validated against independent inspection?

A. Sensor data cannot be stored digitally.

B. A model may identify patterns that do not reliably correspond to actual weld quality.

C. Independent inspection always replaces process monitoring.

D. Machine learning operates only with visual sensors.

Correct Answer: B. A model may identify patterns that do not reliably correspond to actual weld quality.

Question 9

Which statement about collaborative welding robots is most accurate?

A. Their collaborative classification automatically makes arc welding safe without further assessment.

B. They may support human interaction, but the complete welding cell and its hazards still require risk assessment.

C. They do not require guarding, training, or emergency functions.

D. They can safely exceed all force and speed limits during welding.

Correct Answer: B. They may support human interaction, but the complete welding cell and its hazards still require risk assessment.

Question 10

A welding plant wants to improve traceability. Which data association is most useful?

A. Recording only the robot brand and installation date

B. Linking process signals to the part, weld, program version, material, and inspection result

C. Saving only the final average voltage for the entire production shift

D. Recording alarms without identifying the affected component

Correct Answer: B. Linking process signals to the part, weld, program version, material, and inspection result