Robotic finishing automation can make a programmed operation consistent, but repeatability alone does not ensure that a part meets its drawing, finish, or functional requirements. The difference becomes clear in a conventional fabrication example.
On a gas-turbine center casing, full-penetration welds passed ultrasonic testing (UT) against the customer-approved acceptance criteria. After machining, fluorescent penetrant inspection (FPI) revealed fine gas pores open to the surface. The pores had been within the approved UT limits, but machining exposed them and they did not meet the applicable surface-inspection criteria. The response included revising the UT acceptance criteria and adjusting welding-procedure variables to improve shielding of the weld pool.
This case came from conventional fabrication, not a robotic cell. It illustrates a risk relevant to automated manufacturing: a part can pass one stage and fail a later check, and an automated process can repeat the same condition across a batch. Quality must be defined across the complete process, not inferred from successful execution at one stage.
Repeatable Motion Does Not Guarantee Product Quality
Robot repeatability describes how consistently a robot returns to a programmed position. Accuracy describes how close that position is to the intended target. The National Institute of Standards and Technology (NIST) distinguishes the two and explains why both influence a robot’s suitability for a task. Neither, on its own, establishes that the finished part meets its specification.
In robotic deburring or finishing, the result depends on the relationship between the robot path, tool-center point (TCP), fixture, workpiece, and process settings. An unverified TCP after a tool change or collision can misalign the tool path. A casting that seats differently in a fixture can alter tool engagement. A worn abrasive can remove material at a different rate even when the program is unchanged. These conditions can leave a burr, remove too much material, or alter an edge that must meet a defined requirement.
A completed cycle with no machine fault confirms that the sequence ran; it does not confirm that the product is acceptable. Successful motion is an input to quality, not evidence of the final result.
Define the Required Result Before Designing the Process
Before selecting the robot, tooling, fixture, or sensing system, define what an acceptable result means. Drawings, customer requirements, and the component’s function should be translated into criteria that identify the characteristic, its acceptance limit, and the method used to assess it.
For deburring, a note such as “deburr all edges” may express intent but leave room for different interpretations. A usable criterion could specify an edge-break range, a maximum permissible burr, or a defined visual condition checked at an agreed magnification and lighting level against a reference sample. The actual limit must come from the component’s design and functional requirements; there is no universal burr limit for every part. ISO 13715:2017 provides conventions for indicating and dimensioning edges of undefined shape, but it does not establish one acceptance threshold for all deburring operations.
Surface preparation presents a similar distinction. Cleanliness and surface profile are separate characteristics. ISO 8501-1:2007 addresses visual assessment of rust and preparation grades on steel substrates, while ISO 8503-1:2012 addresses surface-profile comparators for abrasive blast-cleaned steel. A surface can meet a cleanliness grade while failing to provide the profile required by the coating system. Specify and verify each characteristic separately.
Inspection priorities should reflect function, not tolerance size alone. A dimension with a relatively broad tolerance may govern fit or assembly, while a tighter dimension may have less influence on function. Engineering should identify features that affect assembly and service and ensure the inspection plan covers them before production begins.
Translate Quality Requirements and Process Variation into Engineering Decisions
Quality requirements should drive decisions about tooling, fixturing, process parameters, and sensing. Common sources of variation include incoming geometry and material condition, part location, fixture wear, abrasive condition, contact force, speed, feed, and dwell time. Identify which inputs affect the final characteristic and decide whether to reduce the variation, accommodate it within a validated range, or measure it and correct the process.
Variation can be reduced upstream by improving consistency in castings or prior machining. Within the cell, stable workholding, suitable tooling, compliant contact, or force control can help manage part-to-part differences. Vision or probing can locate a part or measure a feature so the robot can adjust its path. For a related discussion of when adaptive path planning may be useful, see FPA’s article on adaptive robot path planning versus fixed programming. These controls solve different problems: locating a workpiece does not control contact force, and force control does not prove that the finished edge meets its specification.
Tooling and workholding should be designed around the actual component family and the limits the process must hold. FPA describes this work under custom robotic tooling and fixture design. Every accommodation strategy also needs a defined operating window. If a part falls outside the range the cell was designed and validated to handle, stop the cycle and obtain inspection or engineering disposition. Do not keep compensating until the robot completes the program. A process study should establish which variations the cell can handle, which require correction upstream, and what signals trigger a stop or review.
Process sequence matters as well. Liquid penetrant testing detects discontinuities open to the surface. Under some conditions, grinding or other mechanical finishing can smear metal across a discontinuity’s opening and reduce penetrant access. NASA’s PRC-6506, Rev. G, Liquid Penetrant Inspection includes provisions for pre-penetrant surface preparation under specified conditions. The inspection sequence and any required surface preparation must follow the governing specification and approved procedure. The lesson is not that every finishing operation hides defects; rather, sequence can affect detectability. Decide before cell design whether inspection is required before finishing, after it, or at both stages.
This is why a process study should precede equipment selection: it establishes the actual workflow, part variation, constraints, and acceptance requirements. FPA explains the approach in its article on what a robotic cell process study examines and its process engineering and system design service.
Verify the Output Under Representative Production Conditions
Commissioning should demonstrate that the complete process produces acceptable parts, not simply that the robot follows its path. Define the characteristics to measure, inspection methods, acceptance limits, and range of parts and operating conditions represented in the trial. Testing only an ideal part will not reveal how the cell performs across normal production variation. The sample plan should reflect the qualification objective and the consequences of failure, not a universal sample count.
The measurement method must also be capable of revealing the deviation that matters. In a conventional combustion-liner fabrication program, a fixture used to hold bushings during welding was also used to check their position. Welding heat and mechanical stress shifted the fixture. Because the same reference was used for inspection, the misalignment went undetected until the liners reached the turbine and could not be installed. The corrective action was a dedicated inspection gauge, separate from the production fixture, with independent calibration and control.
For a critical characteristic, do not rely on a production fixture as the sole inspection reference when the manufacturing operation can alter that fixture’s geometry. Use a separate gauge or another independently validated measurement method capable of detecting the relevant deviation.
Repeatability and conformity must also be assessed separately. A process can consistently produce parts outside a limit, while a few acceptable samples do not show how it behaves over time. NIST’s guidance on assessing process stability notes that process stability should be established before process capability is assessed; capability estimates from an unstable process can be misleading.
Maintain Quality After the Cell Enters Production
Passing commissioning establishes a baseline, not a permanent guarantee. Abrasives wear, fixture locators loosen, and tool changes or collisions can alter the relationship between tool and part. A production control plan should define which characteristics are checked, when checks occur, what records are retained, and what happens when a result moves outside the acceptance limit. Changes that could affect conformity should trigger review and re-verification before the changed process is released.
The interval between operations matters too. In a conventional surface-preparation operation, a blasted surface passed checks for cleanliness and profile, but coating application was delayed several days. Improper storage and exposure to moisture and contaminants allowed slight rust to develop, and the coating later failed an adhesion test evaluated under ASTM D3359-23. The initial inspection described the surface at that moment; it could not guarantee that the condition would remain acceptable after storage.
The missing controls were protection during storage, the permitted hold time before coating, and the conditions that required reinspection or re-preparation. Automated cleaning and finishing have the same exposure: handling, storage, and delays can affect whether a part remains ready for the next operation. Define the verification point and the conditions needed to preserve the approved state.
Change control is equally important. In a conventional fabrication case, a supplier replaced a specified single-piece laser-cut Inconel screen with two welded halves to save material. Welding distortion left parts oval and out of flat, and a large batch was rejected. A technical query before production would have allowed engineering to evaluate the proposed change. In an automated cell, a different abrasive, tool, program, or parameter can also affect conformity; evaluate the change and re-verify the process before accepting production output.
Machine data, force signals, and tool-condition indicators can provide early warnings, but treat them as evidence of product quality only after their relationship to the required characteristic has been established. Use them to support defined product checks, not to replace those checks by assumption.
Quality Control in Robotic Finishing: A Pre-Production Checklist
Before approving a robotic finishing cell for production, the team should be able to answer these questions:
- Acceptance criteria: Which characteristics must the finished part meet (such as edge condition, material removal, surface finish, cleanliness, or profile), and what limit applies to each?
- Inspection method and sequence: How will each characteristic be measured, against what reference, and at which process stage? Could finishing affect what a later inspection can detect?
- Part variation: What range of incoming geometry, material condition, and part location is the cell designed and validated to handle? What happens outside that range?
- Commissioning evidence: Do the trial parts and operating conditions represent normal production rather than ideal samples only?
- Process window: What acceptable ranges apply to tool or abrasive condition, contact force, speed, feed, dwell time, and fixture location, and where are they recorded?
- Independent verification: Is the inspection reference independent of production tooling when the manufacturing operation could change the fixture’s geometry?
- Change control and response: Which tool changes, collisions, program edits, fixture changes, or out-of-limit results require a stop, engineering review, containment, or re-verification?
- Downstream protection: How long can a finished surface wait before the next operation, under which storage conditions, and when must it be reinspected or re-prepared?
Build Automation Around Verifiable Quality
Successful automation requires more than a robot that performs a cycle consistently. Manufacturers must define the required result in measurable terms, understand the variation that can affect it, select controls matched to that variation, and verify the finished product with a method capable of detecting unacceptable outcomes.
For manufacturers evaluating robotic deburring, polishing, sanding, or surface preparation, a quality-first assessment can clarify the required finish, sources of variation, and methods needed to verify output. Explore FPA’s robotic finishing automation and process engineering and system design services, or contact the FPA team to discuss a specific application.
This article is general industry information and does not replace applicable specifications, approved procedures, safety requirements, or a qualified engineer’s assessment of a specific application.
About the Author
Chris Urban is the Founder of Forged Path Automation. His 26+ year manufacturing career spans from an international manufacturing specialist trained in Zurich, Switzerland, to corporate President and business owner. Before launching Forged Path Automation (FPA), Chris scaled an industrial gas turbine business unit from its infancy to $50M in value, directed the zero-downtime relocation of 100+ industrial machines to a 150,000 sq. ft. Center of Excellence, and led US operations for a $2.3B global firm. Today, Chris leverages his deep technical roots and an MBA to engineer turnkey robotic finishing cells that deliver total production stability and clear ROI for high-mix manufacturers. Chris holds an advanced background in both the technical and financial sides of manufacturing, combining studies in Applied Science with a Master of Business Administration.
Connect with Chris on LinkedIn to talk shop or discuss your floor’s ROI.
Follow Forged Path Automation on LinkedIn or visit ForgedPathAutomation.com.
