Motivation: The Why?

Currently, the manufacturing industry and tertiary buildings (hotels, hospitals, offices, airports, etc.) are immersed in the fourth industrial revolution, known as Industry 4.0 (I4.0) . This implies a migration towards spaces that operate and are maintained intelligently; in other words, that function effectively and efficiently , and that also optimize and adapt in real time.

The goal is to enable these spaces to fulfill their intended purpose. We can call this smart cognitive buildings and industries 4.0 .

In the case of industry, they are specifically designed to: * Manage complexity. * Produce goods more efficiently. * Be less vulnerable to failure.

A key component in achieving this is the concept of Cyber-Physical Systems (CPS) , which is based on merging the physical world with its digital representation. Therefore, capturing the physical world using IoT sensors and field protocol data is fundamental to determining the current state of a process (spaces, interconnected systems, machines, final control elements, and their environment).

In this way, the data can be used by Generative and Discriminative Artificial Intelligence (AI) methods for more informed cognitive decision-making in cyberspace and, consequently, for better actions in physical space.

What to expect?

As a result, significant improvements are expected. For example, in the area of maintenance, through the scheduling of service activities based on failure forecasts and condition estimates , rather than fixed time intervals or ad hoc actions triggered by simple sensor thresholds.

Production systems developed for this purpose—often called Cyber-Physical Production Systems (CPPS) —are becoming increasingly complex due to: 1. The variety of field systems and protocols (open and closed). 2. Structured and unstructured data. 3. Embedded components and the multitude of interactions between them.

It is crucial that the entity that cooperates and co-maintains understands the context of the entire system .

Unlike manually defining the relationship between parameters, the availability of comprehensive data in Industry 4.0 has shifted the focus to machine learning . We seek to learn the relationship between sensor patterns (input) and the progression of failures (output) to estimate the current state and predict the future.

These approaches are called Data-Driven Predictive Maintenance (PredM) , and their benefits include: * Reduced risk of unplanned downtime. * Increased equipment uptime. * Reduced maintenance costs.

Technically, these approaches typically use waterfall architectures , starting with deep learning models for discriminative tasks and then progressing to generation models with cognitive capabilities. The goal is to uncover knowledge (e.g., degradation indicators) hidden in high-dimensional temporal and/or spatial patterns.

The Challenge in Physical Space

The development of successful ML models (as in computer vision or NLP) is linked to the availability of vast amounts of training examples. However, in industrial applications, success is often only limited in its transferability due to domain-specific properties.

In particular, for PredM , the number of examples representing failures and breakdowns (FaFs) is usually very small due to the high quality of current machines and existing maintenance strategies.

The Solution: Hybrid Integration

To address this challenge, the integration and combination of Local AI Agents with expert knowledge along with data-driven Machine Learning (ML) or Deep Learning (DL) is being considered.

With this approach we achieve the following: 1. Addressing the specific difficulties of the domain. 2. Integrating the knowledge necessary for better diagnosis and early detection. 3. Overcoming the scarcity of examples of flawed training methods.

To do this, we use the Case-Based Reasoning (CBR) methodology, which provides the framework for combining the expert knowledge used by AI Agents and machine learning.

And Edgeman?

This is where Edgemant (Local Cognitive Operation and Maintenance — Edge — for Industry and Buildings) comes in.

Edgemant revolutionizes the maintenance and operation of elevators and solenoid valves with edge AI. It detects faults before they occur and operates/optimizes your equipment, processes, and operations in real time and locally.

How does it work?

What is Edgemant? Edgemant is a suite of AI Agents that operate locally, without internet access , that cooperate autonomously and co-maintain processes and physical equipment in industries and tertiary buildings, based on the prediction and detection of anomalies.

Ready for the next level?

👉 Schedule a visit or try the demo here: https://edgemant.conauti.com/es/