An RVI inspection does not always end when the camera comes out of the tube. The inspector may still have images and video to review, findings to match with notes, locations to confirm, and a report to finish. On a larger tube inspection, those small tasks can add up quickly and create a second round of work after the field inspection is already complete.
Edgevision™ is designed to keep more of that work connected. Its Report Generator and Image Analysis configurations bring inspection capture, AI-assisted review, coding, and report generation into the same field workflow. The goal is not to remove the inspector from the process. It is to make it easier to move from what the camera sees to a finding that is reviewed, organized, and ready for the final report.
Edgevision uses AI-assisted anomaly detection to help inspectors review captured tube imagery and identify areas that may need a closer look. Supported configurations also keep anomaly coding and report generation closer to the inspection, reducing the amount of information that has to be sorted and rebuilt later. The inspector still reviews the imagery, applies the inspection procedure, and makes the final judgment.
AI Vision for Heat Exchanger Tube Sheet & Bundle Inspection
Heat exchanger inspection becomes harder to organize as the number of tubes increases. Many tube interiors can look very similar on camera, especially when the inspection is moving through the bundle one tube after another. A useful image is only part of the record. The team also needs to know where that image came from and what was observed there.
This is why tube position, captured imagery, and the finding should stay connected. A tube map can give the inspection team a common reference for that location. If a finding is associated with a position such as row 2, column 5, the image and inspection notes can be tied back to that same location instead of depending on memory or a folder full of similar-looking files.
Edgevision supports AI-assisted anomaly flagging, coding, and reporting. The row-and-column layout below shows one way those findings could be organized across a sample tube bundle. It is an illustration of the inspection workflow, not a claim that Edgevision automatically creates tube-sheet coordinates.
This 8 × 6 layout is a sample 48-position bundle used only to explain finding organization. Actual heat exchanger tube counts and layouts depend on the equipment being inspected.
Important: the coordinate labels in this illustration show one way an inspection team can organize tube positions. Edgevision's published product information does not state that the system automatically counts tubes or creates row-and-column coordinates.
Pitting, erosion, and crack-like indications can also look different from one inspection to another. Lighting, deposits, surface condition, camera distance, and viewing angle all affect what appears in the image. AI can help direct attention to a potential anomaly, but the inspector still has to review that area, confirm what is visible, and decide what action the inspection procedure requires.
The Manual Inspection Problem
Once the inspection begins producing dozens or hundreds of images, the challenge is keeping the context attached to each one. A frame may clearly show a surface condition, but that image becomes much less useful if its tube position, inspection note, or finding category is stored somewhere else. The inspector may remember those details while the job is active, but rebuilding that context later takes additional time and creates more room for mistakes.
Those problems all come from the same break in the workflow. The inspection captures useful information, but that information has to pass through several separate steps before it becomes a finished record. AI-assisted review becomes more useful when it helps keep those steps connected instead of adding another isolated tool to the process.
Why AI Visual Inspection Matters After the Camera Finds Something
AI-assisted visual inspection starts with the captured image, but its practical value continues after the image is recorded. A potential anomaly still needs to be reviewed, associated with the right location, classified in a consistent way, and carried into the inspection report. When those actions happen in different systems, the inspector may handle the same finding several times.
Edgevision brings those parts of the workflow closer together. Image Analysis can assist with anomaly detection and coding, while the reporting configurations can keep reviewed findings connected to a consolidated job report. The inspector remains responsible for interpreting the image, but less of the surrounding work has to be repeated later.
Where Traditional RVI Reporting Loses Time
Traditional reporting often adds several small handoffs after the field work. The inspector transfers the media, searches for the important frames again, matches images with notes and locations, applies the right finding descriptions, and then places that information into a report. None of those tasks is unusually difficult by itself, but a finding-heavy inspection can turn them into a substantial amount of post-job work.
Keeping review, coding, and reporting closer to the inspection reduces the amount of context that has to be reconstructed. That is the workflow difference the comparison below is intended to illustrate.
A Better Way to Look at Inspection Time
The amount of reporting time on a real job depends on the inspection scope, the number of findings, the customer's reporting requirements, and the way the inspection team already works. Edgevision does not publish a universal time-saving percentage or a guaranteed number of hours, so there is no responsible single number that can be applied to every inspection.
The example below should therefore be read as a workflow illustration rather than a performance claim. Its purpose is to show how the reporting portion of a job can become shorter when review, verification, coding, and report creation are handled with fewer separate steps.
Where Edgevision Can Reduce Post-Inspection Handling
Illustrative reporting-time comparison. Each colored section represents one stage and is sized according to its share of the total time.
These times are an illustrative example, not a guaranteed result or measured Edgevision performance claim. Actual inspection and reporting time depends on the application, number of findings, procedure, operator, and reporting requirements.
What AI-Assisted RVI Means in Edgevision
The shorter workflow shown above depends on what the software is helping the inspector do. In Edgevision, the Image Analysis configuration adds AI anomaly detection, anomaly coding, and AI-assisted anomaly flagging and tagging. These functions are intended to help the inspector work through captured visual data and bring potential areas of interest forward for review.
A flagged image is still the beginning of the inspection decision, not the end of it. The inspector reviews the indication, considers the surrounding surface and image quality, and then applies the inspection procedure or acceptance criteria. If the area needs another NDT method or engineering review, that decision still comes from the inspection process rather than from the AI alone.
AI-assisted defect recognition works best as a review aid. Surface condition, lighting, deposits, camera position, geometry, and image quality can all change what an indication looks like. The inspector should review the flagged area in context before deciding how it should be documented.
What AI Does and Does Not Do in an RVI Inspection
Keeping that distinction clear makes the workflow easier to understand. AI can help an inspector find, organize, and document visual information, but it does not replace the inspection procedure or the judgment required to interpret what the camera sees.
AI can assist with
- Flagging potential visual anomalies for review
- Helping inspectors work through captured imagery
- Supporting anomaly coding and organization
- Keeping findings connected to the reporting workflow
- Reducing repetitive review and documentation steps
AI does not independently
- Decide whether equipment is fit for service
- Replace the inspector or applicable inspection procedure
- Set acceptance criteria
- Determine whether repair or replacement is required
- Replace other NDT methods where subsurface or dimensional data is needed
That division of responsibility is especially important in field RVI because inspection conditions are not perfectly controlled. The software can help the inspector focus attention and keep the record organized, while the human review provides the context needed to decide what a visible indication means for that specific inspection.
How AI-Assisted Defect Recognition Fits Into the Workflow
Once the inspector's role and the AI's role are separated clearly, the workflow becomes straightforward. The camera captures the visual evidence, the software can help surface potential anomalies, and the inspector verifies what should be recorded. The reviewed finding can then stay connected to its coding and report instead of being recreated later.
The advantage of this sequence is continuity. The finding can move from capture to review and then into the report without losing the connection between the image, the inspector's decision, and the inspection record.
On-Device Report Generation Is the Second Half of the AI Story
AI-assisted review can reduce some of the work involved in finding and organizing potential anomalies, but a completed inspection still needs a usable report. If the inspector has to export everything, rebuild the job on another computer, and manually place each image into a separate template, much of the workflow remains disconnected.
Edgevision's Report Generator configuration adds the ability to assemble and generate a consolidated job report with harmonized coding. The Image Analysis configuration includes those reporting functions and adds AI anomaly detection, anomaly coding, and AI-assisted flagging and tagging. This keeps the review and reporting stages closer together.
USB export remains available when inspection files need to be transferred or archived elsewhere. The difference is that file export does not have to be the starting point for building every report. Supported configurations can begin that reporting process on the same field system used during the inspection.
Why Edge Computing Matters for AI Inspection
Keeping these functions close to the inspection also depends on where the processing happens. Edgevision uses an NVIDIA Jetson "Edge" computer to support the integrated video recorder, image capture, and AI-assisted report generation. The system can therefore handle important parts of the inspection workflow locally instead of making a remote cloud connection the center of the job.
That matters in industrial environments where internet access may be limited, unreliable, or restricted. An inspector working around a boiler, heat exchanger, outage area, or plant does not always have the same network access available in an office. Local processing allows the inspection, review, and supported reporting functions to remain available at the jobsite.
The system also supports Wi-Fi control through the Edgevision tablet, supervisory viewing through Wi-Fi and HDMI, and USB data export. Together, those options give the inspection team several ways to view and move the data while keeping the primary inspection workflow on the field system.
AI Vision Inspection Is Different From a Fixed Factory Vision System
The field environment is also why AI-assisted RVI should not be treated like a fixed production-line vision system. A factory camera may inspect the same part from the same position under nearly identical lighting every time. Tube inspection changes continuously as the camera moves through the asset.
Camera distance, angle, reflections, deposits, corrosion products, surface texture, and lighting can all change from one section of tube to the next. The inspector may stop, reverse, adjust the camera position, or revisit an area to understand what the first view showed. Those changing conditions make human review a necessary part of the workflow.
AIT's current Edgevision documentation describes AI-assisted anomaly detection and NVIDIA Jetson edge processing, but it does not identify the underlying AI model architecture. For that reason, Edgevision should not be described as a deep learning system unless that architecture is separately confirmed.
Why This Matters for Boiler and Heat Exchanger Tube RVI
That combination of repetitive imagery, changing inspection conditions, and a large amount of documentation is especially relevant to boiler and heat exchanger tube work. An inspector may move through many similar tube runs while still needing to keep each useful image and finding tied to the right inspection record.
AI-assisted anomaly review can help focus attention during that process, while coding and on-device reporting help carry the reviewed finding forward. The benefit is not that AI performs the inspection by itself. The benefit is that the information collected by the inspector can move through fewer disconnected steps before it reaches the final report.
This article focuses on the AI and reporting side of that workflow. For the camera hardware, tube access, inspection geometry, and other application considerations, see Boiler and Heat Exchanger Tube Inspection: Why Purpose-Built Push Camera Matters.
Compare the Three Edgevision Configurations
The amount of software support needed depends on how the inspection team already works. Some teams mainly need reliable video and still-image capture. Others want the reporting process to stay on the field system, while finding-heavy inspections may benefit from AI-assisted anomaly review in addition to reporting.
Edgevision is available in three configurations so those functions can be matched to the job instead of forcing every inspection team into the same workflow.
| Feature | Basic Video capture | Report Generator Consolidated reports | Image Analysis AI-assisted review |
|---|---|---|---|
| Core System | |||
| HDTV camera system | ✓ | ✓ | ✓ |
| Wi-Fi Edgevision™ tablet | ✓ | ✓ | ✓ |
| Standard accessories | ✓ | ✓ | ✓ |
| Capture and Output | |||
| Video recording | ✓ | ✓ | ✓ |
| Still image capture | ✓ | ✓ | ✓ |
| Flash drive output | ✓ | ✓ | ✓ |
| Reporting and AI Review | |||
| Report generation | - | ✓ | ✓ |
| Harmonized coding | - | ✓ | ✓ |
| Anomaly coding | - | - | ✓ |
| AI anomaly detection | - | - | ✓ |
Illustrative Edgevision Workflows
The configuration table shows what each version includes, but the choice becomes easier when those features are connected to a real inspection workflow. The examples below show how Edgevision's documented reporting and image-analysis functions could fit common tube-inspection programs. They are examples of how the features may be used, not published customer results.
Multi-Site Turnaround Teams
A field service company may send different crews to refinery or petrochemical sites during turnaround season. When several inspectors are producing reports, harmonized coding can give the team a more consistent way to organize findings. On-device report generation can also keep the documentation closer to the work instead of leaving every crew to rebuild reports later.
OEM-Aligned Inspection Programs
An HRSG manufacturer or asset owner may already have inspection procedures, maintenance manuals, and approved terminology for visible findings. Inspectors can use those requirements alongside Edgevision's reporting workflow so the captured imagery, finding descriptions, and final records follow a more consistent process.
Review Across Several Facilities
A reliability team responsible for several plants may receive inspection reports from different crews, sites, or contractors. Consistent coding and report structure can make those records easier to review side by side, especially when the team needs to follow recurring tube conditions over multiple inspections.
What to Check Before Choosing an AI Inspection Camera
Once the workflow is clear, the product still has to match the physical inspection. AI and reporting features do not solve an access problem, and they cannot make up for a camera system that does not fit the tube path or provide usable imagery. The inspection route should therefore be checked alongside the software requirements.
If the inspection equipment category itself is still being decided, AIT's Remote Visual Inspection Tool Selection Guide compares the main types of RVI equipment and explains where each one fits. That broader comparison can help determine whether a push-camera platform such as Edgevision is the right starting point for the application.
The Real Productivity Gain Is Fewer Repeated Steps
There is no single time-saving percentage that describes every Edgevision inspection. Tube count, surface condition, number of findings, inspector workflow, customer requirements, and the amount of reporting all change from one job to the next. A fixed productivity claim would ignore those differences.
The more useful way to look at the workflow is to count how many times the inspection information has to be handled. AI-assisted review can help inspectors work through captured imagery. Harmonized coding can keep confirmed findings organized. On-device reporting can reduce the need to export everything first, rebuild the inspection context, and manually assemble the same information again on another computer.
That is where Edgevision can make the workflow more efficient. The camera still captures the inspection, and the inspector still makes the judgment, but the information collected during the job has a shorter path from the tube to the finished inspection record.
If the application needs a smaller-diameter borescope instead of a push-camera platform, the Mentor Visual iQ+ (MViQ+) is another option with onboard documentation tools and supported AI analytics. AIT can help compare the camera type, probe diameter, length, imaging requirements, and software package against the actual inspection route.
Learn more about Advanced Inspection Technologies or contact AIT to discuss the tube geometry, inspection distance, expected conditions, reporting needs, and the level of AI assistance that makes sense for the job.
