Percepto has introduced Axion, a software platform that analyses industrial inspection data from supported third-party drones, sensors and existing records, extending its technology beyond customers using the company’s own autonomous aircraft.
The platform is available as a subscription service and uses more than 160 specialised AI models developed from over 100,000 hours of autonomous inspection data collected during the past decade, according to Percepto. The figures describe the company’s model portfolio and training experience rather than an independently established rate of defect detection.
Axion brings visual and other inspection records together around individual infrastructure assets. Instead of leaving images, videos and measurements in separate collections, it associates findings with equipment locations, component information and previous inspections.
Electricity networks, oil and gas installations, and industrial sites produce large inspection archives that require review before potential failures or maintenance needs can be identified. Under Percepto’s software-first model, those records can be analysed without first acquiring its Ranger drones or Harbor docking stations.
Compatibility still depends on supported data formats and appropriate interfaces. A platform described as hardware-independent cannot necessarily ingest every file produced by every device without preparation, configuration or integration work.
The technical challenge is partly one of interpretation. A photograph may reveal an unusual shape or damaged component, but an inspection system also needs to determine which asset is pictured, how its parts fit together and whether the observation differs from expected condition.
A conductor that appears disconnected from the correct supporting structure, for instance, may warrant investigation in an electricity network. The significance of that observation depends on recognising the surrounding equipment rather than merely detecting a line-shaped object in an image.
Thermal inspections introduce other variables. Infrared cameras measure radiation from observed surfaces, and converting that signal into an assessment of equipment condition requires suitable assumptions about emissivity, distance, environment and the operating load of the asset.
A hot component is not necessarily defective. The expected operating temperature can vary according to equipment design, electrical current, ambient conditions and the thermal characteristics of its surroundings. A credible assessment needs to account for those factors before creating a maintenance recommendation.
Percepto says Axion uses inspection context and historical records to evaluate such findings. Its product documentation describes computer vision models for component identification and defect detection, thermal analysis, optical gas imaging, image segmentation and language-based exploration of inspection information.
The platform can associate repeated inspections with geographically referenced assets, allowing operators to compare observations made at different times. Consistent asset identification is essential because an apparent trend is unreliable if images from different pieces of equipment are mistakenly combined.
The quality of the historical archive also matters. A gap in inspection records may obscure when deterioration began, while incorrect component data could cause an algorithm to apply an unsuitable reference condition. Centralising information does not itself correct incomplete or inaccurate records.
Axion’s natural-language interface enables personnel to question available inspection findings and asset histories without manually searching every report. Its answers are nevertheless limited by the information held in the system and the accuracy of the underlying analysis.
Percepto describes its reports as expert-validated, with inspection specialists reviewing AI findings. That process differs from allowing a model to authorise physical maintenance or safety-critical intervention independently. It also requires clarity about which observations have been verified and what subsequent examination is needed.
False positive alerts can create unnecessary site visits, work orders and equipment isolation. Missed defects can leave genuine faults unaddressed. The balance between these outcomes depends on the application, inspection quality and the thresholds chosen for detecting anomalies.
Such trade-offs become more demanding as operators expand from occasional surveys to frequent automated monitoring. More data can improve the opportunity to observe changes, but the review workload can grow unless records and alerts are organised around assets and prioritised appropriately.
Existing enterprise asset management systems provide another interface. Maintenance teams may already manage equipment identifiers, failure histories and planned activities within software used across the business. Percepto’s product information refers to APIs and integration options intended to connect Axion findings with those established workflows.
A defect report must still be linked to the correct asset and interpreted within applicable operating procedures. In a highly regulated or hazardous facility, an automated recommendation cannot bypass the approvals and isolation rules governing work on physical equipment.
Percepto has separately secured a multi-year contract with an unnamed US energy company to provide aerial data collection and AI-supported inspection analytics for transmission lines. The arrangement extends an existing relationship, but the company has not disclosed the contract value, number of assets covered or demonstrated reduction in faults.
Its autonomous equipment remains part of the offering. Ranger drones and Harbor stations can collect additional inspection records where customers require more frequent coverage, while existing fleets and other data sources can contribute information under the new software model.
The company is also consolidating engineering, manufacturing and assembly activity in the United States. That operational change is separate from the launch of Axion, and the announcement does not establish a complete withdrawal from all other international operations.
Existing inspection archives can now be processed through Axion without an initial commitment to Percepto’s data collection hardware. The operational test will be whether correctly linked, verified findings can improve maintenance decisions and asset visibility without overwhelming technical teams with additional alerts.
