Today, different industries and tertiary buildings face the challenge of becoming spaces where increasingly efficient products or experiences are created with fewer errors.
The big question is: How can we integrate the knowledge of our engineers (experts) with Machine Learning models that analyze sensors in real time, using past cases to improve failure prediction?
For Edgemant, the answer translates into three layers of technical execution that process the plant's reality from multiple dimensions. We unify what the machine "says" (protocols), what the AI "sees and hears" (audio and video), and what the manufacturer "wrote" (technical literature).
1. The Multimodal and Intellectual Knowledge Graph (Semantic KG)
At Edgemant, the Graph doesn't just map assets; it's the connective tissue of all structured and unstructured data.
-
Total Integration: We link the video streams from camera 04, the decibels from microphone 02, and the IoT sensors (vibration, flow, pressure) directly to the asset in the digital plane.
-
Documentary Context: Each motor is "anchored" to its data sheets, user manuals, and technical PDFs. If a Modbus protocol reports an amperage spike, the graph instantly "knows" that this motor is driving Belt X, knows its operating limits according to the manufacturer, and knows which camera should focus on the component. It's no longer an isolated piece of data; it's pure situational awareness.
2. Deep Learning Engines and IoT Ingestion (The Senses and Nerves)
Edgemant processes massive data streams in real time using three simultaneous analysis paths:
-
Field and IoT Protocols (Nerves): We capture the internal state via Modbus, BACnet, KNX and wireless IoT sensors (LoraWAN/Zigbee).
-
Artificial Vision and Audio (Senses): We apply neural networks to detect visual (color changes due to heat, smoke) and acoustic (friction, cavitation) anomalies before a temperature sensor is even activated.
-
AI Orchestration Agents (Data Fusion): The AI detects that the vibration increased by 10%, but by cross-referencing it with the audio (metallic squeal) and the video (slight smoke), it confirms that it is not a simple reading error, but an imminent failure.
3. Cognitive Reasoning (RAG) and CBR Methodology (The Brain)
This is where Edgemant becomes invincible. The 4R cycle of the CBR (Case-Based Reasoning) methodology is enhanced because our "cases" are now multimedia and multiprotocol:
-
Retrieve: When faced with an unusual pattern, the AI searches for: "When did we have Modbus Error 05 + high-pitched audio at 4kHz + oil leak image?"
-
Reuse: "Six months ago, the camera detected a similar stain; the protocol indicated low pressure, and the expert replaced the hydraulic seal." If this happened at the Lurín plant, the Piura plant already knows how to automatically fix it.
-
Review: The technician receives on their mobile phone the video of the previous fault and the reading of the protocol to apply and validate the repair in minutes.
-
Retain: Edgemant saves this "Master Case". Next time, detection will be instant and 100% accurate.
Furthermore, when a completely new anomaly arises, our AI Agents run RAG (Retrieval-Augmented Generation) . If an "E-102" error is detected on a drive, the AI instantly "reads" the PDF manual and tells the technician: "The manual indicates that E-102 is a phase failure; check terminal 3 as suggested by the manufacturer."
The Differentiating Value: The "Triple Validation"
The thesis is fulfilled with absolute rigor: the combination of Expert Knowledge + Multimodal Deep Learning + CBR is the only way to eliminate false positives and guarantee that the operation never stops.
Edgemant offers unprecedented accuracy because:
-
Validated by Data (Time Series/IoT/Protocols).
-
Valid by Senses (Video/Audio).
-
Valid by Literature (RAG to Manuals/Web).
The result is an accurate and actionable diagnosis that protects the operational continuity of the industry. We've gone from selling "sensor graphs" to delivering an Expert Operating System.
Cyber-Physical Systems (CPS) and Cognitive Understanding
A key component in achieving this technological marvel is the concept of Cyber-Physical Systems (CPS) : the fusion of the physical world with its digital representation.
This means capturing the physical world using structured data (IoT sensors, speed, encoders, temperature) and unstructured data (images, video, audio, text). When machines understand the why , how , and what of the system in which they operate, we are talking about true cognitive understanding .
In Cyber-Physical Production Systems (CPPS) , workspaces have become so complex and interactive that predicting failures takes too long for a human specialist (too many "if" statements in their head). In contrast, Edgemant's AI Agent suite finds immediate cognitive relationships between what it sees, what it hears, historical data, and operating manuals.
The Definitive Approach to Maintenance
To understand why this is revolutionary, we must differentiate Fault (the wear or physical error of the component) from Failure (the failure in the behavior of the system due to the fault ).
Historically, the industry has used two approaches:
-
Model-Based: Formalizes expert knowledge through rules and thresholds. It is robust, but very limited when faced with the difficulty of modeling complex interactions in a CPPS.
-
Data-Driven: Learns from complex relationships in historical data. Its major barrier is the immense amount of specific knowledge that needs to be integrated to model the entire system.
Why does Edgemant outperform them?
Our AI Agents break down these barriers by integrating both perspectives using the CBR methodology. They have three unique capabilities:
-
Learn from existing knowledge.
-
Discovering knowledge that does not yet exist.
-
Orchestrating that knowledge through workflows.
Edgemant's CBR system integrates expert knowledge, machine learning, deep learning, and generative AI not only to detect faults but also to predict failures with an accuracy unattainable by traditional models. This is how we guarantee operational continuity (through co-maintenance, co-operation, and co-supervision), reducing costs, eliminating unplanned downtime, and optimizing the efficiency of the entire physical space.
📞 Ready to take the next step?
Contact us to schedule a demo and/or a visit.
-
📧 Email: edgemant@conauti.com
-
💬 WhatsApp Latam: +51 997975161
-
🇺🇸 WhatsApp USA: +1 (415) 604-6499
-
🤖 Replikers: Chat with Enrique Suárez
-
📋 Form: Complete the form
