Cooling system for a distributed, modular and scalable data processing centre

Abstract

The present invention relates to a distributed, modular and scalable AI processing system comprising at least one portable modular device. Each device is equipped with cooling by immersion in biodegradable dielectric fluid with artificial intelligence controls, uninterruptible power supply units, dehumidifiers and industrial communication modules for native connectivity to Operational Technology (OT). The system executes digital twin models based on physics-informed neural networks (PINN) for energy optimisation.

Description

MODULAR AND SCALABLE DISTRIBUTED DATA PROCESSING CENTER COOLING SYSTEM

[0001] The present invention falls within the scope of high-performance modular processing systems, uninterruptible power supply and communication for Edge data center infrastructure, artificial intelligence applications, and distributed control architectures with remote monitoring in industrial and commercial environments.

[0002] The invention encompasses modular containment systems designed to house interchangeable and interconnectable accelerated processing units, uninterruptible power supply (UPS) units, and communication units. Each system incorporates advanced dielectric fluid immersion cooling, humidity control, and physical protection to ensure stable operation under demanding conditions.

[0003] The rapid development of artificial intelligence (AI) and high-performance computing (HPC) systems has generated an unprecedented demand for technological infrastructures that combine scalability, energy efficiency, and environmental sustainability. This exponential growth in data generation and processing has highlighted the limitations of existing technological paradigms and driven the urgent need for robust solutions that reduce environmental impact and meet increasingly complex operational requirements.

[0004] Historically, data management and computational processing have relied on two main paradigms: traditional centralized data centers and cloud computing services.

[0005] Centralized data centers have formed the backbone of technological infrastructure for decades, concentrating computing resources, storage capacity, and network services in large-scale facilities. However, the exponential growth of data processing has revealed critical limitations in this model:

[0006] -Intensive energy consumption: Running advanced algorithms and deep learning tasks requires the massive use of graphics processing units (GPUs) and other accelerators, which demand considerable amounts of energy.

[0007] -Dependence on water resources: Traditional, water-based cooling systems generate a high environmental impact, increase operating costs and present significant challenges in water-scarce regions.

[0008] -Problems arising from centralization: Concentrating large energy loads in a single location requires complex electricity supply infrastructure, increasing installation and operating costs. This configuration makes it difficult to guarantee a continuous and reliable supply, especially in areas with limitations in the electricity grid, reducing operational flexibility and representing a considerable obstacle to sustainable expansion.

[0009] On the other hand, cloud computing services have emerged as an alternative to outsourcing data processing and storage to remote infrastructures. While this model offers scalability and reduced initial hardware investments, it presents significant challenges:

[0010] -Privacy and security concerns: Transferring sensitive data to remote servers creates security and confidentiality risks, especially critical in sectors such as health, finance, defense, and industrial research.

[0011] -Latency in critical applications: Applications requiring real-time processing, such as industrial control systems, autonomous vehicles, or locally generated data analysis, are affected by the inherent delays in data transfer to remote centers and the receipt of responses. This end-to-end latency is particularly critical in applications demanding rapid decisions, such as safety, industrial process monitoring, and operator assistance.

[0012] -Connectivity and bandwidth dependency: Continuous operation requires stable internet connections, which can be problematic in locations with limited communications infrastructure. Particularly in the context of multimodal data, such as high-resolution video and audio streams, continuous transmission to a remote data center demands considerable bandwidth, which is costly and limiting in environments with restricted or intermittent connectivity.

[0013] There is a growing demand for solutions that allow for the local processing of sensitive data, ensuring that the information remains within the user's premises, thus eliminating the risks associated with external transfer.

[0014] Recently, technical developments such as the decentralized training method proposed by 0G Labs in collaboration with China Mobile have demonstrated the possibility of training AI models with over 100 billion parameters without requiring ultra-high-speed internet or expensive centralized infrastructure. This decentralized approach partially breaks the dependence on centralized cloud platforms and allows data to be kept locally, within the jurisdiction, or even on edge devices, creating new opportunities to reduce costs and accelerate large-scale AI development. However, these proposals do not address the need for modular, transportable, and efficiently cooled physical infrastructure at the point of data generation.

[0015] Existing solutions lack the flexibility needed for efficient deployment in spaces not specifically designed for IT or in locations near industrial data generation points. Furthermore, they generally do not integrate the following capabilities:

[0016] -Tripartite modularity: Architecture that encompasses hardware, physical format and software, allowing interchangeable accommodation of various types of accelerated processing units (GPU, NPU, AI ASICs, FPGAs) without requiring substantial redesigns, and adaptability to different deployment environments, including non-IT spaces such as parking lots, utility rooms or warehouses, without the need for specialized HVAC infrastructure.

[0017] -Robust and native connectivity with Operational Technologies (OT): Direct interconnection with the industrial world through communication modules and specialized protocols (e.g., Modbus, OPC-UA, WirelessHART, ISA100.11a), allowing real-time data reception from sensors, PLCs and actuators of industrial equipment and high-performance computing directly at the edge.

[0018] -Management of digital twins and energy optimization of external processes: Ability to run digital twin models of the external physical process acting as an "energy autopilot" with predictive capabilities, optimizing total consumption (system + external process), extending the lifespan of components and considering proximity to data, local energy cost and the quality of network links in the distributed orchestration of micro-data centers.

[0019] -Advanced cooling systems: Solutions that eliminate dependence on water and dedicated HVAC infrastructure, minimizing energy consumption through the use of biodegradable dielectric fluids and harnessing the surrounding ambient air as an auxiliary cooling medium.

[0020] -Specific system security: Active measures related to monitoring the dielectric fluid (level sensors, leak detection, containment), detection of anomalous movements and physical access control through biometric authentication.

[0021] -Multimodal data processing at the edge: Ability to receive, process and correlate in real time video, still image and audio streams from a plurality of field-distributed capture devices (fixed IP cameras, PTZ cameras, thermal cameras, drones, bodycams, smart glasses), improving the protection of sensitive data by performing the processing mainly within the controlled environment of the system, reducing exposure to public networks, minimizing latency and decreasing the bandwidth required to the cloud by emitting only compact events, descriptions, summaries or metadata.

[0022] The present invention relates to a distributed, modular, and scalable artificial intelligence processing system, comprising one or more physically and logically interconnected portable modular AI processing units, designed to operate in a decentralized manner in industrial and commercial environments.

[0023] The system provides an integrated solution that combines high-performance processing capabilities through artificial intelligence, advanced cooling by immersion in biodegradable dielectric fluid, uninterrupted power supply and communication with Internet of Things (IoT) and Operational Technology (OT) systems, enabling multimodal data processing in proximity to the generation points.

[0024] Each portable modular unit that is part of the distributed system comprises a base case housing computing hardware with computing and storage capabilities, and a top cover externally connected to the base case. The computing hardware is interconnected with an advanced cooling mechanism using biodegradable dielectric fluid immersion, which employs AI-powered electronic controls to dynamically adjust the temperature via fans and radiators, eliminating reliance on water and optimizing energy consumption.

[0025] The distributed system is characterized by its triple modularity: (i) hardware modularity, allowing the interchangeable housing of various accelerated processing units (GPU, NPU, AI ASIC, FPGA); (ii) physical format modularity through scalable containment systems, allowing its deployment in industrial and non-IT environments without requiring specialized HVAC infrastructure; and (iii) software modularity through adaptive processing instructions, incorporating a processor-executed orchestration layer for dynamic workload allocation.

[0026] The modular unit includes dehumidifiers for humidity protection, a structural anchoring mechanism for stability, a drainage system for fluid maintenance, and card handles for easy component removal and insertion. The top cover houses a screen with a biometric authentication system for access control and identity verification, and an AI-powered screen for temperature regulation and monitoring.

[0027] The modular system integrates uninterruptible power supply (UPS) units that provide continuous backup power to ensure operational stability, and industrial communication modules that manage internal and external connections for information exchange, allowing native connectivity with Operational Technology (OT) systems and direct interconnection with industrial sensors, PLCs and actuators through specialized wired and wireless protocols.

[0028] The system is capable of running physics-informed neural network (PINN)-based digital twin models of the external physical process, acting as an energy autopilot that optimizes total consumption, and can interact with integrated Large Language Models (LLM) for advanced analysis and optimization.

[0029] The system incorporates specific safety measures, including level sensors, leak detectors, quick-closing valve actuators, anomalous movement detection, and automatic fire detection and suppression through continuous thermal monitoring. Physical access control is achieved through biometric authentication.

[0030] The system also features a plug-and-play design that facilitates immediate installation through quick-connect interfaces, automatic component recognition, labeled modular cabling, and self-configuration protocols. It is adaptable for both indoor and outdoor installation, with integrated protection against moisture and harsh environmental conditions.

[0031] The invention can also operate individually as an autonomous unit, in a modular fashion with physical interconnection between multiple units forming distributed micro-data centers, or through virtual orchestration defined by processor-executed instructions for the coordinated management of computational resources. This decentralized operation allows data processing close to the points of generation, reducing latency and improving information privacy.

[0032] Therefore, the main object of the invention is to provide a modular containment system for high-performance processing units that overcomes the limitations of centralized data centers and cloud computing, offering a decentralized, energy-efficient and environmentally sustainable solution capable of performing intensive data processing and artificial intelligence in proximity to the data generation points.

[0033] In this context, there is a pressing need for technological infrastructures that combine: scalability and intensive processing capacity; energy efficiency and environmental sustainability; proximity to data generation points; security and local control of information; physical and functional modularity for rapid deployment; and native integration with industrial (OT) environments.

[0034] The present invention is positioned as a solution that addresses these problems through a modular and scalable distributed system for information processing using at least one portable modular artificial intelligence device with improved energy efficiency.

[0035] It provides a modular and scalable architecture system that can accommodate between 2 and 20 high-performance processing units, and from 1 to 5 uninterruptible power supply (UPS) and communication systems, adapting to various configurations and operational demands through attachable and expandable mounting structures that facilitate the formation of decentralized networks.

[0036] It allows for the interchangeable housing of various accelerated processing units (GPU, NPU, AI ASIC, FPGA) through unified anchoring structures, cabling, power supply and cooling, without requiring substantial redesigns of the containment system.

[0037] It facilitates portability and rapid deployment through a modular and assembleable design that allows for efficient transport and installation of the system in different locations, including industrial and non-IT spaces, without the need for dedicated HVAC infrastructure, bringing processing capacity closer to end users and locally generated data.

[0038] It optimizes energy efficiency and thermal management through an advanced immersion cooling system in biodegradable dielectric fluid and utilization of surrounding ambient air, eliminating dependence on water resources and significantly reducing energy consumption compared to traditional cooling systems, with electronic controls that dynamically adjust the temperature using artificial intelligence.

[0039] It integrates an orchestration layer using processor-executed instructions that dynamically assigns artificial intelligence models to accelerators and manages workloads locally or in a distributed manner, optimizing the use of computational resources.

[0040] It provides the system with native connectivity with industrial (OT) and Internet of Things (IoT) systems through specialized communication modules and wired and wireless telecommunications protocols for the acquisition, real-time data processing and generation of control signals for critical processes.

[0041] It allows the execution of digital twin models based on physically informed neural networks (PINN) of the external physical process, acting as an energy autopilot to optimize total consumption (system + external process), avoid idle infrastructure and extend the lifespan of components, through communication with integrated large language models (LLM).

[0042] It provides the ability to receive, process, and correlate in real time streams of documents, video, still images, and audio from a plurality of electronic capture devices distributed in the field, including but not limited to: fixed IP cameras; motorized PTZ or 360° coverage cameras remotely controlled in pan, tilt, and zoom by means of control messages generated by the system; fixed or portable thermal or infrared cameras for monitoring rotating equipment or other critical elements; cameras mounted on unmanned ground or aerial vehicles (drones) that supply visible and / or thermal video streams along with associated telemetry; portable "bodycam" type cameras integrated into helmets or body mounts of operational personnel; and capture devices integrated into glasses, smart helmets, or other personal protective equipment that incorporate microphones for capturing ambient audio and / or voice commands.

[0043] It simplifies installation and maintenance operations through: plug-and-play quick-connect interfaces; self-configuration software and labeled modular cabling; dual anchoring systems for positioning containment units; removable components via telescopic rails; card handles for handling components such as GPUs; inclined tank design with valve for easy draining of dielectric fluid; and remote monitoring capabilities that reduce the need for on-site interventions.

[0044] It guarantees high performance in data processing and execution of artificial intelligence tasks with low latency, operating in proximity to data generation points to ensure real-time processing, simultaneously improving privacy and security by keeping information processed locally, and minimizing the bandwidth required to the cloud by emitting only compact events, descriptions, summaries or metadata.

[0045] It allows the orchestration of multiple systems as a distributed micro-data center, optimizing the use of computing resources, energy consumption profiles and load localization, reducing dependence on large-scale centralized data centers and ensuring operational resilience to network failures, allowing the system to continue analyzing video, audio and process data and assisting operators even with degraded or interrupted external links.

[0046] It incorporates specific safety measures for the dielectric fluid, including level sensors, leak detectors, quick-closing valve actuators, tilt detection and abnormal movements, along with physical access control through biometric authentication and water protection to ensure the durability of the equipment.

[0047] It minimizes environmental impact through a sustainable design that reduces energy consumption, eliminates dependence on water for cooling, uses biodegradable fluids, and maximizes equipment durability through protection against adverse environmental conditions.

[0048] It facilitates the formation of decentralized networks through efficient physical and logical interconnections between multiple systems by means of wired and wireless telecommunications protocols, allowing horizontal scalability and the distribution of computational processing.

[0049] The present invention will be better understood with reference to the accompanying figures, which illustrate, by way of non-limiting example, embodiments of the system of the invention. In these figures, the reference codes indicate identical or similar elements. Fig.1

[0050] It illustrates the internal and external components of the portable modular AI processing system, highlighting elements that provide modularity, adaptability, and efficiency. These include: surveillance cameras, temperature sensors, humidity sensors, an energy consumption meter, ventilation grille modules, a biometric lock, and modular support structures. Fig.2

[0051] This illustrates the internal structure of the portable modular AI processing system, highlighting key components and functionalities. These include: modular AI processing systems, thermal imaging cameras, containment systems for high-performance processing units, support structures configured to house multiple units, containment systems for a telecommunications unit, containment systems for an uninterruptible power supply (UPS), a tray for capturing dielectric fluid leaks, an electric generator motor, and a height-adjustable floor anchoring system. Fig. 3

[0052] It shows a view of the configuration of the portable modular AI processing system, highlighting key components for ensuring reliable fire detection. The identified elements are: thermal imaging cameras, containment systems for high-performance processing units, containment systems for the telecommunications unit, containment systems for the uninterruptible power supply unit, the electric generator, and fire suppression systems. Fig.4

[0053] It shows the support structures of the portable modular AI processing system, highlighting the double anchoring mechanism for the containment systems that support the processing units, consisting of a shaft and a stabilizing arm designed to facilitate maintenance operations. Fig. 5

[0054] It highlights the system's support, surveillance, and physical security elements. These include: thermal imaging cameras for thermal monitoring, a fire suppression system, and a fire extinguishing system with integrated nozzles for active protection. Fig. 6

[0055] It provides a view of the complete AI processing system, incorporating the portable modular AI processing system and its integration with external infrastructure. Key components include: application servers for managing the modular AI system that integrate a video and audio gateway system to normalize multimodal video and audio streams from multiple heterogeneous devices; structured database servers; unstructured database servers; data backup servers; internet connection or telecommunications protocols; Internet of Things (IoT) devices; desktop user workstations; portable user workstations; mobile communications devices; mobile communications and control devices; and Operational Technology (OT) appliances. Fig. 7

[0056] It presents a diagram illustrating the supply and distribution of energy within the complete AI processing system, highlighting the relationship between the different energy sources (electrical grid, UPS, generator) and the subsystems that depend on them for continuous operation. Fig. 8

[0057] It presents a diagram that represents the flow of data exchange through telecommunications protocols within the complete AI processing system, detailing how communication, processing, and field devices are interconnected through wired and wireless protocols to ensure efficient connectivity and distributed operation. Fig. 9

[0058] It illustrates a perspective view of the modular system in an open configuration, showing the main internal components including: top cover, base case, card handles, processing units (GPUs), power supplies, dehumidifiers, control board, central processing unit (CPU), RAM memory, computing hardware, dielectric coolant and insulator, anchoring system with structure, heat sink, holding handle, drainage system and stainless steel tank. Fig. 10

[0059] It illustrates a front perspective view of the modular system in closed configuration, showing the user interfaces including: screen with biometric authentication system, screen with artificial intelligence controls for dynamic temperature regulation, dehumidifiers and base box. Fig. 11

[0060] It illustrates a rear perspective view of the modular system, showing the thermal management components including: fans, containment units with immersion cooling system, adaptive electronic controls with artificial intelligence, external radiators for heat dissipation and drainage system. List of elements of the invention

[0061] The following list corresponds to the set of elements that are part of the present invention. At the front of each element, the reference code included in the set of figures is indicated for easy identification. GENERAL SYSTEM Modular Processing System for AI 10 Portable Modular AI Processing Equipment 20 Distributed, Modular, Scalable AI Processing System 30 System Control and Monitoring Center 40 Language Modeling Platform 50 Management, Control, and Monitoring Platform 60 MONITORING AND SECURITY COMPONENTS Surveillance Cameras 100 Temperature Sensors 101 Humidity Sensors 102 Energy Consumption Meter 103 Grid Modules 104 Biometric Lock 105 Internal Support Structure 106 External Support Structure 107 STRUCTURAL AND CONTAINMENT COMPONENTS Thermal Imaging Cameras 200 Containment Systems for High-Performance Processing Units 201 Support Structures to House Processing Units 202 Containment Systems for UnitTelecommunications 203 Containment systems for uninterruptible power supply unit 204 Oil leak capture tray 205 Electric generating motor 206 Height-adjustable floor anchoring system 207 FIRE PROTECTION SYSTEMS Fire suppression systems 300 Support structures with double anchoring mechanism 400 Fire extinguishing system with integrated nozzles 500 EXTERNAL INFRASTRUCTURE AND COMMUNICATIONS Server with application for IA modular system management 600 Structured database servers 601 Unstructured database servers 602 Information backup servers 603 Internet or telecommunications protocol 604 IoT device 605 Desktop user station 606 Portable user station 607 Mobile communications device 608 Mobile communications and control device 609 OT device 610 Video and audio gateway system 611 Ethernet 614 INTERNAL COMPONENTS MODULAR TABLET HOLDER Top cover 21 Boxbase22Card Handles700GPU Enclosures701Power Supplies702Dehumidifiers703Computer Hardware704Board714Motherboard724CPU734RAM744Storage System754Communications System764Coolant and Insulator (Dielectric Fluid)705Heat Sink706Grip Handles707Drainage System708Container Tank709Structure Anchoring Mechanism710USER AND CONTROL INTERFACESDisplay with Biometric Authentication System800Display with AI for Temperature Regulation801ADVANCED COOLING SYSTEMFans810Containment Units with Immersion Cooling System811Adaptive Electronic Controls with AI812Radiators813Sensors814

[0062] This disclosure pertains to a detailed explanation of some embodiments of the invention and does not imply any limitation of the scope of the invention as defined in the claims. For a more complete understanding of the present invention, reference is made to the following illustrative and non-limiting description, taken in conjunction with the accompanying drawings, in which the reference codes assigned to each part indicate identical or similar elements in the corresponding drawings.

[0063] In general terms, the modular and scalable distributed AI processing system (30) is a set of multiple electronic devices, computers or AI information processing components, interconnected in a network, that work together on AI processing tasks, and that can easily grow by adding more components without affecting the operation of the overall system.

[0064] The system (30) operates in a distributed manner through physical or logical connections between its components, following a modular architecture that includes hardware elements, control software for application servers and databases, as well as processing modules with AI algorithms. The system (30) receives information from multiple sources, specifically IoT devices, OT devices, workstations, and mobile devices, through various telecommunications protocols.

[0065] The modular and scalable distributed AI processing system (30) incorporates one or more high-performance portable modular AI processing systems (10). In turn, each system (10) protects and integrates the portable modular AI processing equipment (20), which constitutes the information processing core of both the systems (30) and the (10).

[0066] This system (30) incorporates the following subsystems: a control and monitoring subsystem, which monitors and verifies the status of the system and the different components and exchange of information; a telecommunications subsystem, which allows the exchange of information between different components of the system through wired and wireless telecommunications protocols; a processing subsystem, which through software platforms and algorithms that incorporate artificial intelligence allows data processing, even in a generative way; a power supply subsystem, which keeps the system equipment running, delivering the currents and voltages necessary for its proper functioning.

[0067] Portable modular high-performance AI processing system

[0068] The high-performance portable modular AI processing system (10) is a next-generation technology solution that integrates efficiency, scalability, and sustainability to address the challenges of data-intensive processing and artificial intelligence in diverse deployment environments, including industrial facilities, healthcare centers, and commercial buildings. Its architecture is optimized to ensure continuous and efficient performance in any scenario.

[0069] The system (10) forms the processing core of the modular and scalable distributed AI processing system (30). The systems (10) are designed to execute complex artificial intelligence tasks with high efficiency and reliability, housing modular containment systems that act as external and internal support structures. Integrating the system (10) into modular support structures allows for rapid installation in approximately two hours, along with customization capabilities, scalability, and ease of transport.

[0070] The system (10) incorporates uninterruptible power supply (UPS) units that provide continuous backup power to ensure operational stability, as well as communication units that manage the exchange of information through telecommunications protocols via physical or logical connections, both internal and external.

[0071] The portable, high-performance AI processing modular system (10) is adaptable for indoor or outdoor installations, with water resistance providing protection against humidity and environmental conditions. The systems (10) can be interconnected to form decentralized processing and communication networks, enabling load balancing, operational redundancy, and resource optimization.

[0072] The portable, high-performance AI processing modular systems (10) contain cooling systems that employ a biodegradable dielectric medium, with dielectric oil being the preferred implementation method. The recirculation system comprises pumps that draw the dielectric medium and send it to the radiators, operating in a redundant configuration to ensure operational continuity. The hydraulic circuit incorporates check valves that regulate the unidirectional flow of the medium. The radiators, installed in the rear and front sections of the system, have additional units as backup and integrate fans for forced heat dissipation.

[0073] Thermal management is complemented by integrated heat sinks in both the casing and the lid of the portable modular AI processing equipment (20), whose configuration is adjusted to the specific conditions of the installation site.

[0074] The system incorporates forced convection stirring mechanisms that maintain a uniform and homogeneous temperature in the dielectric medium, preventing the formation of hot spots. It operates with an additional stirrer as a backup. Protective grilles are designed to prevent cables or components from entering the stirrer area.

[0075] Thermal monitoring is performed using four strategically placed temperature sensors: two sensors monitor the temperature at the recirculation system's inlet nozzle, while the other two are located at the top of the dielectric medium. The system integrates intelligent electronic controls based on artificial intelligence algorithms capable of detecting pump and radiator failures and automatically redirecting the flow using solenoid valves that activate redundant components. These controls dynamically regulate the medium's temperature by adjusting the pump flow rate, fan RPM, and forced convection system RPM, based on information provided by both internal and external temperature and humidity sensors.

[0076] The mounting structures are designed to position and support the external and internal components of the high-performance portable modular AI processing system (10). These structures have a detachable configuration to facilitate transport and installation.

[0077] The structural system incorporates a double anchoring mechanism comprising a shaft and a stabilizing arm, allowing each portable modular AI processing unit (20) to be positioned in an ergonomic position that facilitates maintenance operations.

[0078] The structures feature anchors specifically designed to organize different types of cables, including fiber optic, Ethernet, and power cables, enabling efficient and secure connections between the containment systems. The structures are designed with height adjustment to regulate the spacing between the different containment systems according to installation requirements.

[0079] Proper installation is ensured by a circular bubble level that guarantees the system is plumb, complemented by solid anchor holes designed for fixing to both the floor and walls, providing structural stability. The structures incorporate a lower tray designed to contain and collect any potential leaks of the dielectric medium from the system (10), as shown in Figure 2.

[0080] The portable, high-performance AI processing modular systems (10) incorporate integrated security, monitoring, and automated protection systems. The surveillance system comprises surveillance cameras (100) that ensure security by monitoring the surrounding area and providing real-time visual feedback, complemented by thermal imaging cameras designed to monitor both internal and external temperatures of the system, including equipment (20), UPS units, communication systems, and cabling. These thermal imaging cameras integrate artificial intelligence algorithms capable of predicting fire risks through predictive analytics, as illustrated in Figures 1-3 and 5.

[0081] Environmental monitoring is performed using temperature sensors that record the ambient conditions within the system and send this information to the equipment (20) to dynamically adjust its cooling according to the surrounding conditions. Additionally, humidity sensors (102) are strategically located to monitor ambient humidity levels and protect sensitive hardware. Humidity control is further enhanced by dehumidifiers integrated into the equipment (20), UPS units, and communication systems, protecting internal components from moisture.

[0082] Access security is ensured through biometric locks that prevent unauthorized access, confirm the operator's identity, and generate traceability by recording the identification and the time when maintenance operations are performed.

[0083] The system integrates an automated fire suppression system equipped with nozzles that operate in conjunction with detection systems using thermal cameras and artificial intelligence, allowing its automatic orientation towards the point where the fire is being generated, using CO₂ extinguishers as the extinguishing agent in a preferred mode, as shown in figures 3 and 5.

[0084] The management of potential leaks of the dielectric medium is carried out using containment trays and drainage systems designed to capture and manage such leaks. The system (10) allows for remote monitoring for the continuous supervision of the thermal imaging cameras, temperature and humidity sensors, as well as the electronic controls, minimizing the need for on-site maintenance interventions. Portable modular AI processing equipment

[0085] Portable modular AI processing equipment (20) consists of equipment that incorporates the physical mechanical and electronic hardware and software elements necessary to perform computing activities using AI. These systems are the core of information processing, receiving data in various formats—text files, office documents, audio, video, and images—and processing them using software to achieve a specific goal through AI processing algorithms.

[0086] According to Figures 9 to 11, the invention corresponds to a piece of equipment (20) consisting of a modular container that stores and protects computing hardware (704), responsible for processing large volumes of data in real time through instructions executed by a processor that incorporates artificial intelligence.

[0087] The equipment (20), through different configurations, has the possibility of interconnecting physically or logically with other equipment (20) through different wired and wireless telecommunications protocols thanks to the communications system (764), also allowing connection with various IoT devices, such as sensors and devices that transmit information through telecommunications protocols in data, audio and video formats.

[0088] The equipment (20) has an advanced cooling and heat dissipation system, as well as the necessary elements for its easy portability and movement, allowing its placement in small environments and taking advantage of elements of the environment to control its temperature.

[0089] The set of essential characteristics of the object of the modular invention are shown in Figures 9 to 11, viewed from different angles, which allow observation of: an advanced cooling system that minimizes energy consumption through the use of biodegradable dielectric fluid, including redundant pumps and radiators, fans and forced convection systems, as well as temperature and humidity sensors; a modular design that allows for quick installation and customization, facilitating transport and scalability of the system; adaptability that allows the system to adapt to different sizes of motherboards known in the state of the art (ATX, microATX and E-ATX); and a containment system designed with an incline in the tank floor and a valve to facilitate emptying and maintenance, present in the drainage system. Image and audio capture devices

[0090] In one operating mode, the equipment (20) can receive audio and / or video information from selected image and audio capture devices from the group consisting of: a) CCTV-type video surveillance cameras with an IP interface (e.g., compliant with RTSP, ONVIF, or other streaming protocols over Ethernet); b) motorized PTZ or 360° coverage cameras, capable of being remotely controlled in pan, tilt, and zoom by means of control messages generated by the equipment (20); c) thermal or infrared cameras, both fixed and portable, intended for monitoring rotating equipment or other critical elements; d) cameras mounted on unmanned ground or aerial vehicles (e.g., drones), which supply visible and / or thermal video streams along with associated telemetry; e) portable "bodycam"-type cameras integrated into helmets or body mounts of operational personnel;(yf) Capture devices integrated into smart glasses, smart helmets or other personal protective equipment, which also incorporate microphones for capturing ambient audio and / or voice commands.;

[0091] In some embodiments, one or more of the aforementioned devices connect to the equipment (20) via a wired network infrastructure based on industrial Ethernet, for example, through one or more PoE (Power over Ethernet) switches that provide both power and IP connectivity to the cameras. In other embodiments, the connection is established wirelessly, using industrial Wi-Fi access points, dedicated radio links, cellular links (4G / 5G / 6G), or satellite links. In these cases, the field devices transmit their data streams to one or more communication gateways, which, in turn, forward these streams to the equipment (20) via Ethernet or optical links.

[0092] In specific implementations geared towards industrial, mining, oil, or heavy manufacturing environments, the capture devices may be installed in classified or hard-to-reach areas. In these cases, the modular communications architecture of the device (20) allows for the integration, via external modules or gateways, of additional communication technologies (e.g., licensed or unlicensed band radio links, long-range point-to-point links, or LoRaWAN gateways or others), while maintaining a unified interface within the device (20) based on wired or optical IP links.

[0093] Figure 9 shows the incorporation of the communications system (764) into the equipment (20) with the possibility of intercommunicating through the aforementioned protocols, to send and receive data and audio and video information.

[0094] Taken together, the modular integration of video and audio capture devices described allows the equipment (20) to act as a specialized AI edge micro data center, capable of ingesting, normalizing and analyzing locally multimodal flows from industrial cameras, operator handheld devices and mobile platforms, reducing dependence on traditional data center infrastructures and high-capacity network links to the cloud.

[0095] In additional preferred embodiments, artificial intelligence processing is performed in a decentralized manner on each existing machine (20), without the need to continuously send video and audio streams to remote data centers. Equipment subsystems

[0096] The equipment (20) is structured by the following integrated subsystems that operate in a coordinated manner to provide high-performance processing with optimized energy efficiency: Containment subsystem and modular structure

[0097] The containment and modular structure subsystem forms the physical basis of the equipment (20), comprising the base box (22), the container tank (709), preferably made of stainless steel, and the top cover (21). This subsystem integrates the structural anchoring mechanism (710), which allows coupling with other modular equipment (20) to form expandable structures, and houses the adjustable rail system that facilitates the dynamic reconfiguration of internal components. The modular structure not only physically protects the internal components but also defines the spaces for housing the dielectric fluid and provides the mechanical interconnection points between all other subsystems.

[0098] The containment systems are designed to house processing units, UPS and communication units; adaptability to different sizes of motherboards known in the state of the art (ATX, microATX and E-ATX, etc.); containment system designed with an incline in the tank floor and a valve to facilitate emptying and maintenance. Processing subsystem

[0099] Within the containment volume operates the processing subsystem, which constitutes the computational core of the system. This subsystem comprises a GPU cabinet (701) capable of housing multiple interchangeable accelerated processing units (GPU, NPU, AI ASIC, FPGA), the CPU (734), and the computing hardware (704) which integrates a board (714) for system control, the motherboard (724) for interconnection and control of electronic components, RAM memory (744) for volatile data storage, and a storage system (754) for non-volatile data and information storage.

[0100] This subsystem is in close interaction with the advanced cooling subsystem, as all its components are immersed in direct contact with the dielectric fluid, and it receives stable power from the power supply subsystem while executing the instructions coordinated by the artificial intelligence and orchestration subsystems.

[0101] It includes the ability to install and configure different software applications that incorporate AI algorithms such as LLM (Large Language Model), which is a type of Artificial Intelligence designed to understand, generate and manipulate human language in an extremely natural and coherent way.

[0102] The GPU enclosure (701) is adaptable to the physical arrangement of the GPUs that perform data processing, allowing accordion-style configurations with varying physical separation distances between them, depending on the level of data processing required to optimize cooling. This separation is performed intelligently and automatically using instructions executed by a processor that is part of the control layer. Advanced cooling subsystem

[0103] The advanced cooling subsystem maintains optimal thermal conditions for the continuous operation of the processing subsystem. It comprises the refrigerant and insulator made of biodegradable dielectric fluid (705), circulation pumps that move the fluid through the system, radiators (813) that transfer heat to the outside, fans (810) that facilitate heat dissipation, and cooling containment units (811).

[0104] This subsystem operates under the direct supervision of the monitoring, control, and safety subsystem, which, through sensors (814), provides real-time data on temperature, humidity, and fluid level to the AI-powered adaptive electronic controls (812). These controls dynamically adjust the pump flow rate and fan speed. Dehumidifiers (703) housed in the top cover complement this subsystem by protecting the components from ambient humidity, while heat sinks (706) distributed around the structure contribute to passive heat dissipation. Power supply subsystem

[0105] The power supply subsystem provides stable and redundant power to all system components through power supplies (702) and uninterruptible power supply (UPS) systems.

[0106] This subsystem interconnects directly with the processing subsystem to supply the power required by the accelerated processing units and computing hardware, with the cooling subsystem to power the pumps and fans, and with the monitoring, control, and security subsystem to power the sensors, displays, and authentication systems. Redundant power distribution ensures operational continuity even in the event of external power grid failures, and is coordinated by the monitoring subsystem, which supervises energy consumption in real time.

[0107] OT / IoT industrial communications and integration subsystem

[0108] The industrial communications and integration (OT / IoT) subsystem enables system connectivity with the outside world and with industrial devices. It comprises the communications system (764), industrial communication modules that implement wired (Modbus TCP, PROFINET, EtherNet / IP, RS-485 / 422, Industrial Ethernet, OPC-UA) and wireless (Wi-Fi, Bluetooth / BLE, LTE / 5G, NB-IoT, LoRaWAN, WirelessHART, ISA100.11a) protocols, and the control panel (714). This subsystem allows for real-time data acquisition from industrial sensors, PLCs, and external actuators.

[0109] It is functionally interconnected with the artificial intelligence and digital twin subsystem, which processes received industrial data and generates control signals for external equipment, and with the orchestration subsystem when multiple modular systems operate interconnected in a distributed network.

[0110] Additionally, this subsystem facilitates connectivity with the multimedia processing subsystem for receiving video and audio streams from field-distributed capture devices. Monitoring, control and security subsystem

[0111] The monitoring, control, and security subsystem supervises and protects the operation of all other subsystems. It integrates the biometric authentication display (800) that controls physical access to the equipment through fingerprint, iris, face, or voice recognition; the AI-powered temperature control display (801) that provides a user interface for system control; and a network of multiparameter sensors (814) that monitor fluid and component temperature, dielectric fluid level, leak detection, abnormal tilt, and internal humidity.

[0112] This subsystem is interconnected with the refrigeration subsystem, providing it with data for the dynamic adjustment of thermal parameters; with the artificial intelligence subsystem, which processes sensor data for predictive optimization; and with the quick-closing valve actuators, which execute automatic safety procedures when anomalous conditions are detected. The access and operation event log provides complete traceability for auditing. Maintenance and accessibility subsystem

[0113] The maintenance and accessibility subsystem facilitates system service and upgrade operations. It comprises the drainage system (708) with a valve and inclined tank design that allows for complete evacuation of the dielectric fluid, card handles (700) that facilitate the removal and insertion of processing components, gripping handles (707) for safe system transport, telescopic rails for vertical removal of base plates and components, and the double anchoring system for stable positioning during interventions.

[0114] This subsystem is interconnected with the containment subsystem, providing ergonomic access to internal components; with the cooling subsystem, through the drainage system that allows for maintenance of the dielectric fluid; and with the monitoring subsystem, which enables remote diagnostics, reducing the need for on-site interventions. Distributed orchestration and management subsystem

[0115] The orchestration and distributed management subsystem coordinates the operation of multiple modular systems functioning as a distributed micro data center. Implemented through the software orchestration layer, it comprises the task scheduler that dynamically assigns inference and training workloads, the communication optimization module that uses low-communicative distributed training algorithms, the consolidated telemetry system that collects temperature, flow rates, power consumption, and link quality data from each modular system, and multi-tenant billing management for inter-organizational compensation schemes.

[0116] This subsystem communicates with the communications subsystem to exchange telemetry and tasks between distributed systems, with the processing subsystem for the allocation of computational loads, and with the artificial intelligence subsystem for the coordination of distributed models, simultaneously optimizing the use of resources, energy profiles and location of loads with respect to data sources.

[0117] Artificial intelligence subsystem and digital twins

[0118] The artificial intelligence and digital twin subsystem provides advanced optimization and intelligent processing capabilities. It runs physics-informed neural network (PINN)-based digital twin models that represent external physical processes such as furnaces, boilers, heat exchangers, and production lines; large language models (LLMs) that interpret the outputs of the digital twins and produce structured control instructions or recommendations in natural language; and energy optimization algorithms that function as an energy autopilot by adjusting both internal system parameters and operating parameters of coupled industrial equipment.

[0119] This subsystem receives data from external sensors via the communications subsystem, generates control signals including setpoints for the speed of the cooling subsystem's pumps and fans, valve adjustments, and commands that modify the relative position of the GPU modules using the adjustable rail system of the containment subsystem, optimizing the distribution of the dielectric fluid and internal temperature to extend the lifespan of the components. It continuously interacts with the monitoring subsystem to receive internal sensor data and with the orchestration subsystem when operating in a distributed configuration.

[0120] Real-time multimedia processing subsystem

[0121] The real-time multimedia processing subsystem enables the ingestion, analysis, and correlation of video, image, and audio streams from distributed capture devices. It comprises multimedia capture interfaces that receive data from CCTV surveillance cameras with IP interfaces, motorized PTZ cameras, thermal or infrared cameras, cameras mounted on unmanned aerial vehicles (drones), body cameras, and capture devices integrated into smart glasses or smart helmets with microphones for audio; stream processors that execute multimodal correlation algorithms within the processing subsystem; and compression systems that generate compact events, descriptions, summaries, or metadata, drastically reducing the bandwidth required on external networks.

[0122] This subsystem receives multimedia streams through the communications subsystem via industrial Ethernet connections with PoE switches, industrial Wi-Fi, cellular links (4G / 5G / 6G) or satellite links, performs AI processing locally in the processing subsystem avoiding continuous transmission of streams to remote data centers, and can interact with the AI ​​subsystem when multimedia analytics feed digital twin models for industrial or safety process optimization. Team components (20)

[0123] As can be seen so far, the artificial intelligence processing equipment (20) focuses on a containment system designed to house hardware components essential for high-performance processing. A detailed report of the invention's parts, their reference numbering, hardware and software definitions, and their contribution to the invention is presented below: Base box

[0124] As shown in Figure 1, the base case (22) is the structure that protects and houses the system's hardware and cooling components. In a preferred embodiment, the base case (22) is a stainless steel container tank (709) that defines the internal volume of the housing, but the material is not a limiting factor.

[0125] Inside the base box (22) are: power supplies (702), computing hardware (704) and coolant and insulation (705); and outside, the heat sink (706), the holding handles (707), the drainage system (708) and an anchoring mechanism with structure (710). Top lid

[0126] The top cover (21) is the upper part of the system that protects the internal components housed in the base box (22). The following are interconnected within the top cover (21): the base box (22), the display with biometric authentication system (800), the AI ​​display for temperature control (801), and the dehumidifiers (703). Card handles

[0127] Card handles (700) are physical components that allow manipulation of the system's internal cards. They facilitate the removal and insertion of cards, enabling safe maintenance and upgrades. GPU Processing Unit Enclosure

[0128] The GPU processing unit enclosure (701) is the physical section that houses the graphics processing units (GPUs), which are essential for data processing, complex operations, and instruction execution. It enables the high-performance processing required for artificial intelligence and advanced computing applications. Power sources

[0129] Power supplies (702) are units that provide electrical power to the equipment components (20). They ensure a stable, necessary, and redundant supply of power, voltage, and current for the continuous operation of the system. Dehumidifiers

[0130] Dehumidifiers (703) are devices that control the humidity inside the equipment (20). They protect electronic components from moisture damage, ensuring system reliability. Computer hardware

[0131] Computing hardware (704) includes the main processing components, which include: board (714), motherboard (724), CPU (734), RAM (744), storage system (754), and communications system (764). These devices and components provide the computing and storage capacity necessary for data processing and application execution.

[0132] The circuit board (714) refers to a control panel for the operation of the computer hardware (704). The motherboard (724) refers to the electronic circuit board that interconnects the different elements of the computer hardware (704). The CPU (734) is the electronic processing circuit, usually a microprocessor or microcontroller, that executes instructions to perform operations and control the system hardware. RAM (744) refers to random access memory (RAM), which is the main memory where data and programs currently in use are temporarily stored.

[0133] The storage system (754) is a non-volatile data storage device that permanently stores all digital information (operating system, programs, files, etc.) of the computer, even when it is turned off.

[0134] The communications system (764) consists of hardware and software components configured to interoperate directly with field networks and process control (OT) systems, including wired (Modbus TCP, PROFINET, EtherNet / IP, RS-485 / 422, Industrial Ethernet) and wireless (Wi-Fi, Bluetooth / BLE, LTE / 5G, NB-IoT, LoRaWAN, WirelessHART, ISA100.11a) protocols. These components enable real-time data acquisition from industrial sensors and actuators, and the execution of AI algorithms to generate control signals in operating environments close to the data source. Furthermore, they facilitate communication between devices using various wired and wireless communication protocols. Coolant and insulator

[0135] The coolant and insulator (705) is the dielectric fluid used for cooling the internal components. It allows for efficient and safe cooling, minimizing energy consumption and maintaining optimal operating temperatures.

[0136] There are mainly three types of dielectric oils used in the modern electrical industry, classified according to their base composition: mineral dielectric oils derived from petroleum refining, vegetable dielectric oils from renewable vegetable sources, and synthetic dielectric oils generally obtained in the laboratory.

[0137] In a preferred embodiment of the invention, the dielectric heat transfer fluid comprises a synthetic hydrocarbon-based composition, specifically hydrogenated 1-decene dimer (CAS No.: 68649-11-6) at a concentration of approximately 99.7% by weight, with an antioxidant stabilizing additive consisting of branched alkyl 3,5-bis(1,1-dimethylethyl)-4-hydroxybenzene propanoate (CAS No.: 125643-61-0) at a concentration of approximately 0.3% by weight. This formulation results in a dielectric fluid specifically designed for cooling electrical and electronic components.

[0138] The fluid exhibits exceptional physicochemical properties for use in electrical cooling systems. It is characterized as a transparent liquid with a relative density of 0.8 at 20°C and a kinematic viscosity of 8.1 cSt measured at 40°C, facilitating its circulation in immersion cooling systems. The fluid has an extremely wide operating range, with a freezing point of -65°C and an initial boiling point above 350°C, enabling operation in extreme environmental conditions. The n-octanol / water partition coefficient is greater than 4.82 at 20°C, and its water solubility is less than 80 ppm, characteristics that confirm its hydrophobic nature and compatibility with moisture-sensitive electronic components.

[0139] A significant advantage of the described dielectric fluid is its superior safety profile. The flash point, measured according to ASTM D92 (open cup), is 190°C, and the auto-ignition temperature exceeds 350°C, giving it an NFPA flammability rating of 1, which categorizes it as a fire-resistant fluid. The vapor pressure at 20°C is exceptionally low (0.01 kPa), virtually eliminating the risk of vapor inhalation under normal operating conditions. The material is non-explosive and non-oxidizing, and it is chemically stable under normal use, preventing hazardous reactions up to temperatures above 290°C.

[0140] The dielectric fluid exhibits remarkably low toxicity characteristics. Toxicological studies demonstrate an oral LD50 in rats greater than 10,000 mg / kg and a dermal LD50 in rabbits greater than 3,500 mg / kg, classifying it as non-toxic to humans and animals. From an environmental perspective, the fluid is inherently biodegradable according to REACH criteria, with LC50 values ​​for fish (96h) greater than 10,000 mg / l, EC50 for crustaceans (48h) greater than 10,000 mg / l, and EC50 for algae (72h) greater than 1,000 mg / l. The material has no bioaccumulation potential, does not meet the PBT (Persistent, Bioaccumulative and Toxic) or vPvB (Very Persistent and Very Bioaccumulative) criteria according to Annex XIII of the REACH Regulation, and does not have endocrine disrupting properties, representing an environmentally responsible solution for industrial cooling applications. Heat sink

[0141] The heat sink (706) is a component that helps control the heat generated by electronic components. It improves cooling efficiency, extending the lifespan of the components and optimizing performance. Grip handles

[0142] As shown in Figure 9, the gripping handles (707) are external elements located on the front and sides of the base box (22) that allow the system to be held and handled safely. They facilitate the transport and installation of the equipment (20). Drainage system

[0143] The drain system (708) is a mechanism for evacuating the dielectric fluid from the tank. It facilitates maintenance and replacement of the dielectric fluid, simplifying service procedures. Container tank

[0144] The containment tank (709) is a three-dimensional housing that contains the components and the dielectric fluid. It provides robust protection and durability to the containment system. In one embodiment of the invention, the containment tank (709) is made of stainless steel, but this feature is not a limiting factor. Anchoring mechanism with structure

[0145] The anchoring mechanism with structure (710) is an external side mechanism of the base box (22) of the equipment (20) that ensures the equipment is fixed to the supporting mechanical structure. It provides stability and safety, facilitating the integration and transport of the equipment. Screen with biometric authentication system

[0146] As shown in Figure 10, the biometric authentication screen (800) integrates a biometric recognition system for access control. Through instructions executed by a biometric recognition processor, it controls access to the equipment (20). It enables secure access and traceability of maintenance performed, with the ability to verify biometric data such as fingerprints, face, or iris scans. AI-powered display for temperature regulation

[0147] As shown in Figure 10, the AI-powered temperature control display (801) is a screen that displays temperature information and enables regulation via artificial intelligence. AI algorithms that adjust the system temperature allow for precise temperature control, optimizing performance and energy efficiency. Fans

[0148] As shown in Figure 11, the fans (810) are devices that facilitate air circulation for cooling. They improve heat dissipation, helping to maintain an optimal temperature.

[0149] Containment units with cooling system

[0150] Containment units with cooling systems (811) are structures that house and protect components, integrating a cooling system. They ensure the stable operation of the components, optimizing thermal management. AI-powered adaptive electronic controls

[0151] AI-powered adaptive electronic controls (812) are electronic components that control and adapt system operation using processor-executed instructions with artificial intelligence algorithms that optimize system performance. They detect faults in redundant pumps and radiators through a solenoid-valve-based blocking system. Radiators

[0152] Radiators (813) are devices that dissipate heat from the dielectric fluid; they cool the dielectric fluid, improving the cooling efficiency, as shown in Figure 11. Sensors

[0153] Sensors (814) are devices (often called transducers) designed to detect and measure a physical or chemical quantity in the environment (such as temperature, light, pressure, distance, fluid level, or inclination) and convert that information into an electrical signal (voltage, current, or resistance) that can be interpreted and processed by an electronic system. Their essential function is to provide accurate data for monitoring and controlling system parameters, preventing damage, and ensuring operational efficiency. Modular anchoring and interconnection mechanisms

[0154] The modular anchoring and interconnection mechanisms correspond to the set of elements that adapt to the base box (22) and allow for anchoring and securing. This category includes the anchoring mechanism with structure (710) and securing handles (707). Ventilation and drainage system

[0155] The ventilation and drainage system is the set of elements responsible for the cooling of the equipment (20) that incorporate: at least one fan (810), containment units with cooling system (811), electronic adaptation controls with AI (812), radiators (813) and sensors (814). Team interconnections

[0156] The following interconnections are evident in the equipment (20):

[0157] The equipment (20) is tangibly delimited by a base box (22) which, in a preferred embodiment, consists of a container tank (709) preferably made of stainless steel, and a top cover (21) assembled to said base box (22).

[0158] The top cover (21) is assembled to the base box (22) and has the following components attached: some dehumidifiers (703), the screen with biometric authentication system (800) and the AI ​​screen for temperature regulation (801), as shown in figures 9 and 10.

[0159] The heat sink (706) and side mounting handles (707) are located on the front of the base case (22). The mounting mechanism with frame (710) is located on the sides of the base case (22), and the drainage system (708) is located on the lower side section. The fans (810) are assembled in the cooling containment units (811), along with the AI ​​adaptive electronic controls (812), the radiators (813), and the drainage system (708).

[0160] Now, within the interior volume of the base case (22) are housed and interconnected the following components: the computing hardware (704) comprising the circuit board (714) and the motherboard (724). Interconnected to the motherboard (724) are: the CPU (734), the RAM (744), the storage system (754), the communications system (764), and the sensors (814).

[0161] The power supplies (702) are housed and interconnected with the equipment components and the coolant and insulator (705) is housed within the inner volume of the base case (22), in direct contact with the processing components and computing hardware.

[0162] The equipment (20) comprises a stainless steel tank equipped with a lid that integrates gripping handles and an anchoring system to the clamping structure, providing robust protection and durability to the assembly, as shown in figures 2 and 3. This tank houses the computing system that includes a motherboard, CPU, RAM memory and information storage, providing the processing and storage capacity necessary for the execution of AI algorithms.

[0163] The units (20) feature an external motherboard mounting bracket, as shown in Figure 4A, which accommodates various form factors, specifically ATX, microATX, and E-ATX. This bracket includes telescopic rails that facilitate vertical motherboard removal and insertion. Additionally, the units (20) contain a dynamic and intelligent enclosure configured to house GPUs, LPUs, DPUs, and Infiniband and Ethernet communication cards, designed for computationally intensive tasks such as AI training and data analysis. This enclosure has a card mounting system on its underside that allows for adjustment of card sizes and the spacing between cards to optimize layout and power dissipation efficiency. The cards have handles attached to their tops for easy removal and installation.

[0164] The power supply system comprises modular power supplies with their respective anchors and cable ties to ensure reliable operation and redundancy. A telescopic multi-device anchor, connected to the modular power supplies, is designed with telescopic rails for easy removal and insertion. Power distribution is managed by a PDU with its corresponding anchor.

[0165] Thermal management is achieved through a dielectric dissipation medium that, in one configuration, uses biodegradable dielectric oil to efficiently cool components and maintain optimal operating temperatures. The system includes dehumidifiers installed on top of the tank, designed to capture and control any incoming moisture, protecting sensitive electronics. A heat sink complements the thermal management by efficiently dissipating heat from critical components. The system incorporates a drain mechanism for extracting and receiving the dielectric oil during maintenance or replacement operations.

[0166] The anchoring system with its structure secures the containment units within the modular frame, ensuring stability and ease of integration. Externally linked handles on each unit (20) allow for convenient transport and positioning.

[0167] The equipment (20) also integrates a biometric lock that prevents unauthorized access and allows for maintenance traceability, as well as an emergency button that allows for a quick system stop.

[0168] External connectivity is established via waterproof Ethernet female connectors that provide secure connectivity, along with power supply connectors. The entire system design is waterproof and modular.

[0169] These portable modular units (20) incorporate the following subsystems: an external protection and security subsystem, which includes surveillance cameras (100) and biometric sensors to ensure system security and prevent access by unauthorized personnel; an internal protection and security subsystem; an external anchoring subsystem that incorporates the external support structures of the components and elements; an internal anchoring subsystem that incorporates the internal support structures of the components and elements; a fire detection subsystem, which incorporates the fire detection and extinguishing control elements; a processing subsystem, which incorporates the portable modular AI processing systems. Distributed system infrastructure (30)

[0170] The modular and scalable distributed AI processing system (30) integrates an efficient power flow and telecommunications architecture that ensures continuous and secure operations.

[0171] As illustrated in Figure 7, the primary power source is the main electrical grid that supplies the UPS system. A generator provides backup power in case of a main grid failure. The UPS centralizes and manages power distribution to all connected units, including the Automatic Transfer Switches (ATS), Surge Protection Devices (SPDs), and circuit breakers, which distribute power to the high-performance portable modular AI processing systems (10) and telecommunications systems, ensuring uninterrupted operation. The energy consumption meter (103) allows for real-time monitoring of energy consumption for efficient system management.

[0172] As shown in Figure 8, external clients connect to the system (30) via telecommunications protocols, with Ethernet being a representative example. Internet access provides connectivity to external networks. Telecommunications systems, including switches, routers, modems, and firewalls, act as the management core, managing incoming and outgoing connections to ensure they are stable, secure, and efficient. Processed data is distributed to systems (10), ensuring the functionality of AI applications and data analysis using artificial intelligence algorithms.

[0173] The connection can be established wirelessly, using industrial Wi-Fi access points, dedicated radio links, cellular links (4G / 5G / 6G), or satellite links. In these cases, field devices transmit their data streams to one or more communication gateways, which in turn forward these streams to the modular and scalable distributed AI processing system (30) via Ethernet or optical links. In specific implementations for industrial, mining, oil, or heavy manufacturing environments, the system's modular communication architecture (30) allows for the integration of additional communication technologies via external modules or gateways, while maintaining a unified internal interface based on wired or optical IP links.

[0174] The modular and scalable distributed AI processing system (30) is designed as a plug-and-play solution, enabling immediate installation via pre-designed and pre-configured connections. The system incorporates standard quick-connect interfaces that automatically integrate into the system upon physical connection, complemented by labeled modular cabling designed for intuitive connections using universal connectors for both data and power.

[0175] Component recognition is performed automatically by an initialization software platform that adjusts the operating parameters of the modules without requiring manual intervention. This platform works in conjunction with self-configuration protocols capable of synchronizing the system (30) with local networks and external devices, eliminating the need for advanced programming during the installation process.

[0176] The system integrates visual indicators and alerts that verify and confirm the correct installation and operation of the components immediately after connection, providing real-time feedback on the system's operational status.

[0177] The complementary components of the invention are as follows: server infrastructure, system control and monitoring center, language modeling platform, telecommunications infrastructure, field and operating devices, user stations and interaction interfaces, and surveillance and control system.

[0178] The application server for managing the modular AI system (600) is a high-performance computer dedicated to running the software that monitors, controls, and manages the various components of the system (30), including the portable modular AI processing equipment (20), structured database servers, unstructured database servers, and information backup servers, as illustrated in Figure 6. These functions are performed by intercommunication via the Internet or telecommunications protocols, including Ethernet.

[0179] Structured database servers (601) are computers specialized in storing, managing, and accessing highly organized data using SQL (Structured Query Language). Conversely, unstructured database servers (602) are optimized for storing and managing large volumes of data that do not follow a fixed schema, such as documents, images, and sensor data, frequently employing NoSQL systems. Data backup servers (603) are computers dedicated to backing up critical data, ensuring redundancy and recovery in case of data loss.

[0180] The system control and monitoring center (40) is a centralized location or interface from which personnel can supervise, manage, and coordinate all operations, configurations, and alarm notifications of the modular and scalable distributed AI processing system (30). It operates through the management, control, and monitoring platform (60), which is a software platform that enables control and monitoring to manage, activate, deactivate, and verify alerts from the different system components, installed on the server with the modular AI system management application (600).

[0181] The language model platform (50) incorporates software that includes artificial intelligence algorithms and may include platforms such as LLM (Large Language Model) which is a type of Artificial Intelligence designed to understand, generate and manipulate human language in a natural and coherent way.

[0182] The telecommunications network (604) comprises the global network of interconnected computers and the set of protocols, such as TCP / IP, that enable communication and data exchange between systems. Ethernet (614) is a standard local area network (LAN) technology and protocol that enables communication between devices via wired connections.

[0183] IoT devices (605) are electronic physical objects, such as sensors, meters, or cameras, equipped with sensors and software to connect to the Internet and exchange data with other devices and systems. OT devices (610) comprise hardware and software used to monitor and control physical processes, devices, and industrial events, including fire suppression system controllers and temperature controls.

[0184] Desktop user stations (606) are fixed personal computers optimized for use on workstations by operators or end users who interact with the equipment (20). Portable user stations (607) are laptop computers that offer the user mobility to interact with the equipment (20) from different locations.

[0185] Mobile communication devices (608) are handheld devices, such as smartphones or tablets, that enable communication and network access for interacting with equipment (20). Mobile communication and control devices (609) not only enable communication but also have the capability to monitor and control equipment (20) through dedicated software applications. All these devices and their interconnections with the system (30) are shown in Figure 6.

[0186] The Video and Audio Gateway system (611) acts as a bridge or translator device between different networks or protocols, allowing the flow of video and audio data between the surveillance system and the control center.

[0187] Functional subsystems of the distributed system (30)

[0188] The components described above are functionally integrated within the modular and scalable distributed AI processing system (30) through specialized subsystems that operate in a coordinated manner to ensure efficient information processing, operational stability, and comprehensive system management. Each subsystem groups related components that perform specific functions within the modular architecture, enabling distributed, scalable, and resilient operation of the entire system. Control and monitoring subsystem

[0189] The control and monitoring subsystem is responsible for supervising and verifying the operational status of the system (30) and its components, as well as managing the exchange of information between the different elements of the system. As shown in Figure 6, this subsystem incorporates the server with the application for managing the modular AI system (600), the structured database servers (601), the unstructured database servers (602), the information backup servers (603), and the system's control and monitoring center (40), operating in an integrated manner to ensure continuous monitoring and efficient management of all processes within the distributed system. Telecommunications subsystem

[0190] The telecommunications subsystem has the function of allowing the exchange of information between the different components of the system (30) through wired and wireless telecommunications protocols.

[0191] According to Figures 2 and 6, this subsystem incorporates the telecommunications network (604) that provides the overall communication infrastructure, the Ethernet technology (614) that enables high-performance local area connections, and the telecommunications unit containment systems (203) that house and protect the network equipment. These components operate in a coordinated manner to ensure stable, secure, and efficient connectivity, facilitating communication between the equipment (20), servers, and external devices connected to the distributed system. Processing subsystem

[0192] The artificial intelligence processing subsystem has the function of executing data processing, including generative processing, using artificial intelligence software and algorithms.

[0193] Figure 6 shows that this subsystem consists of the modular AI processing systems (10) that provide the hardware and cooling infrastructure, the portable modular AI processing units (20) that house the accelerated computing components using GPUs, LPUs, and DPUs, the video and audio gateway system (611) that manages the multimedia data flow, and the server with the application for managing the modular AI system (600) that monitors and coordinates the processing operations. These components operate in an integrated manner to execute complex data processing, analysis, and generation tasks using high-performance artificial intelligence algorithms. Power supply subsystem

[0194] The power supply subsystem has the function of keeping all the equipment in the system (30) running, providing the currents and voltages necessary for its correct and continuous operation.

[0195] As shown in Figures 1 and 2, this subsystem incorporates the power generation engine (206) that provides backup power in the event of main grid failures, the uninterruptible power supply (UPS) containment system (204) that houses the UPS units responsible for managing and distributing power to all system components, and the energy consumption meter (103) that monitors energy consumption in real time. These components operate in a coordinated manner to ensure a stable power supply, operational redundancy in the event of grid interruptions, and efficient control of the distributed system's energy consumption. External protection and security subsystem

[0196] The external protection and security subsystem is designed to ensure the security of the system (30) and prevent unauthorized access through continuous monitoring and access control. As shown in Figure 1, this subsystem incorporates surveillance cameras (100) that provide real-time visual monitoring of the surrounding area, temperature sensors (101) that monitor ambient thermal conditions, humidity sensors (102) that detect humidity levels that could compromise equipment, an energy consumption meter (103) that records energy consumption to detect operational anomalies, grille modules (104) that provide physical protection and controlled ventilation, and a biometric lock (105) that controls access through biometric authentication and generates traceability of interventions.These components operate in an integrated manner to provide continuous surveillance, monitoring of environmental conditions, physical protection of the infrastructure, and access control to the distributed system. Internal protection and security subsystem

[0197] The internal protection and safety subsystem is designed to ensure the protection of the system's internal components (30) against critical events such as fires and dielectric fluid spills. As shown in Figures 1, 2, and 5, this subsystem incorporates the internal support structure (106) that provides structural integrity and organization of the internal components, the fire suppression systems (300) that detect and respond to thermal hazards, the fire extinguishing system with integrated nozzles (500) that enables automatic and targeted extinguishing through intelligent aiming at the fire's source, and the dielectric fluid leak capture tray (205) that contains and collects any potential spills of the cooling medium.These components operate in an integrated manner to provide structural support, automated fire detection and suppression, as well as safe management of dielectric leaks. External anchoring subsystem

[0198] The external anchoring subsystem has the function of providing the external support structures that hold and organize the system's components and elements (30). As shown in Figure 1, this subsystem incorporates the external support structure (107), which is designed with a detachable configuration to facilitate transport and installation, providing structural stability, mechanical integrity, and modular interconnection capacity between multiple systems (10), also allowing for height adjustment and proper leveling through the elements described in the fastening structures. Internal anchoring subsystem

[0199] The internal anchoring subsystem provides the internal support structures for the system components and elements (30). This subsystem consists of the internal support structure (106), the height-adjustable floor anchoring system (207), and the support structures for housing processing units (202), which operate in an integrated manner to ensure internal organization, structural stability, and adaptability to different equipment configurations. Fire detection subsystem

[0200] The fire detection and suppression subsystem incorporates the elements for early detection and automated fire suppression control within the system (30). As shown in Figures 3 and 5, this subsystem consists of the thermal imaging cameras (200), the fire suppression systems (300), and the fire suppression system with integrated nozzles (500), which operate in an integrated manner to provide continuous thermal monitoring, predictive fire hazard detection, and automated fire suppression. Support and anchoring subsystem

[0201] The support and anchoring subsystem is designed to position and support the external and internal components of the high-performance portable modular AI processing system (10). As shown in Figures 2, 4, and 4A, this subsystem incorporates the support structures with a double-anchoring mechanism (400), comprising a shaft and a stabilizing arm that allow each unit (20) to be positioned ergonomically for maintenance operations; the dielectric fluid leak capture tray (205), which contains and collects any spills of the cooling medium; and the height-adjustable floor anchoring system (207), which allows the systems (10) to be fixed and leveled with vertical adjustment capabilities according to specific installation requirements.These components operate in an integrated manner to provide structural stability, ease of access during maintenance, adaptability to different spatial configurations, and safe management of dielectric medium leakage. Ventilation subsystem

[0202] The ventilation subsystem is designed to provide controlled ventilation and continuous monitoring of the system's environmental conditions (30). Figure 1 shows that this subsystem incorporates grille modules (104) that allow air circulation and provide physical protection while maintaining controlled access to the system's interior, temperature sensors (101) that continuously monitor both internal and external ambient thermal conditions, and humidity sensors (102) that detect ambient humidity levels to prevent condensation and protect sensitive components. These components operate in an integrated manner to ensure adequate air circulation, real-time thermal monitoring, humidity control, and protection of equipment against adverse environmental conditions. System interconnections (30)

[0203] The way in which the essential elements of the invention are interconnected is as follows:

[0204] The modular and scalable distributed AI processing system (30) operates through an architecture of physical and logical interconnections between its components, guaranteeing communication, power supply and comprehensive operational coordination.

[0205] The surveillance cameras (100) are interconnected with the external support structures (107), the processing subsystem, and the power supply subsystem. The temperature sensors (101) and humidity sensors (102) maintain identical interconnections with the external support structures (107), the processing subsystem, and the power supply subsystem. The energy consumption meter (103) is connected to the external support structures (107) and the power supply subsystem. The grid modules (104) are interconnected only with the external support structures (107), while the biometric lock (105) is connected to the external support structures (107) and the power supply subsystem.

[0206] The external support structures (107) constitute the integration point for the surveillance cameras (100), temperature sensors (101), humidity sensors (102), energy consumption meter (103), grid modules (104), and biometric lock (105). Meanwhile, the thermal imaging cameras (200) are interconnected with the internal support structure (106).

[0207] The internal support structure (106) constitutes the central element of interconnection with multiple system components, connecting with the thermal imaging cameras (200), the containment systems for high-performance processing units (201), the support structures to house processing units (202), the containment systems for the telecommunication unit (203), the containment systems for the uninterruptible power supply unit (204), the tray for capturing dielectric fluid leaks (205), the electric generating motor (206), the floor anchoring system with adjustable height (207), the fire suppression system (300), the support structures with double anchoring mechanism (400), and the fire extinguishing systems with integrated nozzles (500).

[0208] The containment systems for high-performance processing units (201) are interconnected with the internal support structure (106), the power generation motor (206), the containment systems for the uninterruptible power supply unit (204), the portable modular AI processing equipment (20), the containment systems for the telecommunication unit (203), the telecommunication network (604), and Ethernet (614). The support structures for housing processing units (202) are connected with the internal support structure (106) and with the portable modular AI processing equipment (20). The support structures with a double-anchoring mechanism (400) are interconnected with the internal support structure (106) and with the portable modular AI processing equipment (20).

[0209] The containment systems for the telecommunications unit (203) are interconnected with the internal support structure (106), the telecommunications network (604), and Ethernet (614). The video and audio gateway system (611) is connected to the portable, high-performance AI processing modular system (10).

[0210] The containment systems for the uninterruptible power supply unit (204) are interconnected with the internal support structure (106). The electric generator (206) constitutes a critical interconnection element, connected to the internal support structure (106), the portable modular AI processing equipment (20), the surveillance cameras (100), the temperature sensors (101), the humidity sensors (102), the energy consumption meter (103), the biometric lock (105), the thermal imaging cameras (200), the containment systems for the telecommunications unit (203), the containment systems for the uninterruptible power supply unit (204), the fire suppression systems (300), and the fire extinguishing system with integrated nozzles (500).

[0211] The fire suppression system (300) and the fire extinguishing system with integrated nozzles (500) are interconnected with the internal support structure (106). The trays for capturing dielectric fluid leaks (205) are linked to the internal support structure (106). The height-adjustable floor anchoring system (207) is part of the internal support structure (106).

[0212] The AI ​​modular system management application server (600) is interconnected with the structured database servers (601), the unstructured database servers (602), the data backup servers (603), the AI ​​modular processing system (10), the system control and monitoring center (40), and the telecommunications network (604). The structured database servers (601), the unstructured database servers (602), and the data backup servers (603) are interconnected with the AI ​​modular system management application server (600). The telecommunications network (604) is connected to the AI ​​modular system management application server (600) and the high-performance portable AI processing modular system (10).

[0213] IoT devices (605), desktop user stations (606), portable user stations (607), mobile communication devices (608), mobile communication and control devices (609), and OT devices (610) are interconnected with the high-performance AI processing modular portable system (10), enabling interaction and data exchange between users, field devices, and the distributed system processing core. Operation

[0214] The modular and scalable distributed AI processing system (30), as a comprehensive technology platform, orchestrates multiple components for the efficient processing of information using artificial intelligence in decentralized environments. Its modular and adaptable design allows for rapid installation and intelligent interconnection, ensuring efficiency, scalability, and sustainability.

[0215] The system (30) is initiated by a power-up process that performs a complete self-diagnostic to ensure that all components are operational and ready to process information. Once the system's integrity is verified, the data acquisition and normalization process begins. Basic equipment operating flow (20)

[0216] The modular device (20) has the following operating modes: Device initialization: The device (20) powers on and performs a self-diagnostic test to ensure all components are operational. Access control: System access is controlled by the biometric authentication system (800), which can verify biometric data such as fingerprints, face, or iris scans. Processing execution: The GPUs (701) and CPU (734) execute artificial intelligence algorithms and data processing tasks. Heat absorption: The coolant and insulator (705) absorbs the heat generated by the processing components. Heat dissipation: The cooling system, composed of fans (810) and radiators (813), dissipates this heat to maintain an optimal temperature. Dynamic monitoring and adjustment: The AI-powered temperature control display (801) monitors the device (20) and adjusts parameters to maximize efficiency and security.Redundancy management and fault detection: The AI-powered adaptive electronic controls (812) detect pump and radiator faults and redirect fluid flow using redundant pumps and radiators via a solenoid-valve-based blocking system. Maintenance: The drain system (708) facilitates maintenance and replacement of the dielectric fluid. a. Data Acquisition and Normalization (Processing and Telecommunications Subsystem).

[0217] IoT devices (605) and OT devices (610), such as industrial sensors, PLCs, and actuators, capture real-time data from their operating environment. Simultaneously, surveillance cameras (100), temperature sensors (101), humidity sensors (102), and thermal imaging cameras (200), located on external support structures (107), collect multimodal streams of video, still images, audio, and environmental readings, as shown in Figures 1 and 2.

[0218] The video and audio gateway system (611) receives video streams from multiple heterogeneous devices, normalizes them by decoding and recoding formats, adjusting resolution and frame rate, and generating image sequences suitable for local AI processing, as shown in Figure 6. The telecommunications containment system (203) manages incoming and outgoing connections, acting as the core of the telecommunications subsystem. This system receives data via wired network infrastructure, such as industrial Ethernet, or wirelessly via Wi-Fi, LTE / 5G, or satellite links from field devices, relaying it to the system (30) via Ethernet (614) or optical links, as shown in Figures 6 and 8.The telecommunications network (604) allows data exchange between system components and with external networks, including Ethernet clients, ensuring stable and secure connectivity as shown in Figure 6.

[0219] b. Data Processing and Artificial Intelligence (Processing Subsystem)

[0220] The portable modular AI processing units (20) integrated into the high-performance portable modular AI processing system (10) constitute the processing core, housing computing hardware including CPU, RAM, and data storage. The accelerated processing units (GPUs, LPUs, DPUs), and communication cards contained within the units (20) execute advanced artificial intelligence algorithms and data-intensive processing tasks, as shown in Figures 1, 2, 3, and 6.

[0221] The server with the application for managing the modular AI system (600) and the software orchestration layer dynamically assign AI models to the accelerators and manage workloads locally or in a distributed manner. The system can run digital twin models based on physically informed neural networks (PINNs) of the external physical process, acting as an automatic energy optimization system that extends the lifespan of components by communicating with integrated Large Language Models (LLMs), as shown in Figure 6. AI processing is primarily performed in the modules (20) and systems (10) close to the data source, minimizing latency and bandwidth required to the cloud by transmitting only compact events, descriptions, or metadata.

[0222] c. Cooling and Thermal Management (Support and Anchoring Subsystem and Ventilation Subsystem)

[0223] The biodegradable dielectric medium, specifically dielectric oil, contained in the stainless steel tank within the system (10), absorbs the heat generated by the high-performance processing components, including GPUs and CPUs, as shown in Figures 4 and 5. The cooling system, consisting of pumps, radiators, fans, and forced convection stirring systems, dissipates the heat from the dielectric medium to maintain optimal operating temperatures, as shown in Figures 3, 4, and 5.

[0224] Intelligent electronic controls based on AI algorithms detect pump and radiator faults, redirect oil flow using redundant components via solenoid valves, and regulate temperature by adjusting pump flow rates and fan and agitator RPMs. This control is based on information provided by internal and external temperature (101) and humidity (102) sensors, as shown in Figure 4. Grille modules (104) allow the entry of ambient air as an auxiliary cooling medium, optimizing energy efficiency, as shown in Figure 1.

[0225] d. Energy Supply and Management (Energy Supply Subsystem)

[0226] The main electrical grid constitutes the primary power source, supplying the containment system for the uninterruptible power supply (204). The electric generator (206) provides emergency or supplementary power in the event of a mains power failure, ensuring operational continuity. This generator is interconnected with various elements, including systems (10), surveillance cameras (100), temperature sensors (101), humidity sensors (102), energy consumption meter (103), biometric lock (105), thermal imaging cameras (200), containment system for the telecommunications unit (203), containment systems for the uninterruptible power supply (204), fire suppression systems (300), and a fire extinguishing system with integrated nozzles (500).

[0227] The uninterruptible power supply (UPS) containment systems (204) centralize and manage power distribution to all connected systems, integrating protection mechanisms such as ATS, SPDs, and circuit breakers to ensure power stability. The energy consumption meter (103) tracks the system's energy usage, enabling efficient consumption management and optimization.

[0228] e. System Monitoring and Control (Control and Monitoring Subsystem)

[0229] Temperature sensors (101) and humidity sensors (102) continuously monitor internal and external environmental conditions to ensure operation within optimal parameters, as shown in Figures 1 and 2. Thermal imaging cameras (200) identify critical temperature changes in real time, improving safety and thermal efficiency by preventing overheating and predicting fire risks using artificial intelligence algorithms, as shown in Figures 1 and 2.

[0230] The system control and monitoring center (40) is the centralized location from which personnel supervise, manage, and coordinate all system operations, configurations, and alarm notifications (30), interconnecting with the application server for managing the modular AI system (600). This server runs the software that monitors, controls, and manages the various system components, including the portable modular AI processing equipment (20), structured database servers (601), unstructured database servers (602), and data backup servers (603), via the telecommunications subsystem. f. Automated Security and Protection

[0231] Biometric locks (105) prevent unauthorized access, confirm operator identity, and generate traceability of maintenance activities, as shown in Figures 1 and 2. The fire suppression systems (300) and the fire extinguishing system with integrated nozzles (500) use thermal imaging cameras (200) and artificial intelligence to automatically detect and direct the nozzles to the exact point of an incident, effectively suppressing fires, as shown in Figures 3, 4, and 5. The dielectric fluid leak capture tray (205), located beneath the units, collects any potential leaks from the cooling system, ensuring environmental safety. g. Maintenance and Adaptability

[0232] The drainage system facilitates the removal of the dielectric medium for maintenance or replacement, simplifying service processes. The support structures with a double-anchoring mechanism (400) include a shaft and a stabilizing arm that allow the containment units to be positioned ergonomically for maintenance, facilitating access to internal components, as shown in Figures 4 and 4A. Card handles facilitate the removal and insertion of GPUs and other processing cards for upgrades or repairs within the equipment (20). Remote monitoring significantly reduces the need for on-site maintenance by continuously monitoring system conditions using thermal imaging cameras and distributed sensors.

[0233] This comprehensive operational flow enables the modular and scalable distributed AI processing system (30) to operate autonomously and efficiently, adapting to diverse operational needs and ensuring robust, secure, and sustainable AI processing in decentralized environments. Specific equipment modalities (20)

[0234] The system's decentralized and compact design allows for efficient implementation in a wide variety of environments. Some of its applications include factories, communication towers, hospitals, shopping malls and hotels, financial institutions, and building parking garages.

[0235] According to the interconnections between the systems, they can operate in the following way: individual, modular with physical connection (wiring), modular with logical connection (communication protocols), virtual (software defined).

[0236] The following seven specific equipment modalities are detailed (20):

[0237] Triple modularity (hardware, physical format and software)

[0238] In this configuration, the containment system is configured to interchangeably house different types of accelerated processing units, including one or more of: graphics processing units (GPUs), neural processing units (NPUs), application-specific integrated circuits for artificial intelligence (AI ASICs), embedded edge computing modules and / or programmable gate arrays (FPGAs), so that the same mechanical anchoring structures, electrical connection and fluid circulation support replacements and expansions without substantial redesign of the tank.

[0239] This modularity extends to selected equipment format variants from cabinet-type unit, rack module, floor unit or pod, which share a principle of immersion in dielectric fluid and heat dissipation by external radiators.

[0240] Native integration with IoT / industrial (OT) systems

[0241] In another modality, the equipment (20) comprises one or more industrial communication modules configured to interoperate with field networks and process control systems, including wired and wireless industrial buses and automation protocols.

[0242] The communication modules include, for example, interfaces for industrial Ethernet, industrial serial buses, high-level protocols such as OPC-UA, as well as wireless interfaces based on Wi-Fi, Bluetooth, cellular networks (including LTE and 5G) and industrial wireless standards such as WirelessHART and / or ISA100.11a.

[0243] The modules allow the system to receive real-time data from external sensors and actuators, and execute, through instructions stored in memory and executed by one or more artificial intelligence processors, algorithms that analyze said data and generate control signals intended to adjust operating parameters of coupled industrial equipment.

[0244] Digital twin / "energy autopilot" of the external process

[0245] In an advanced embodiment, the system runs one or more of its own digital twin models, configurable to represent different external physical processes, such as ovens, boilers, heat exchangers, production lines, or other industrial equipment.

[0246] Each digital twin model is implemented as a surrogate model based on Physics-Informed Neural Networks (PINN) neural networks, trained from operational data and known physical relationships of the process, configured to predict the thermal and energy behavior of the process and the containment system.

[0247] PINN-based digital twin models receive, through industrial communication modules, data from sensors associated with the physical process and generate, through processor-executed instructions, control signals and / or adjustment parameters intended to optimize the energy consumption of the external physical process and the system itself, including the reduction of idle infrastructure by consolidating workloads and shutting down or transitioning to low-power states of underutilized computing resources.

[0248] The control signals include, for example, setpoints for the speed of dielectric fluid recirculation pumps, setpoints for the speed of radiator fans, hydraulic circuit valve settings, and commands to actuators that modify the relative position of GPU modules mounted on an internal rail system, so as to optimize fluid distribution and temperature inside the tank, helping to increase the lifespan of the GPUs.

[0249] At least one large language model (LLM) running on the system communicates with the PINN models of the digital twin to interpret their outputs, combine them with additional contextual information, and produce structured control instructions and / or recommendations in natural language, so that the same digital twin technology embedded in the system can be reused in multiple energy-intensive industries.

[0250] In a preferred mode, the equipment (20) acts as an energy autopilot that adjusts both internal cooling parameters of the containment system and operating parameters of the coupled industrial equipment.

[0251] Installation in "non-IT" spaces without the need for dedicated HVAC

[0252] Due to the airtight containment of the dielectric fluid and the heat dissipation by means of external radiators coupled to a closed hydraulic circuit, the equipment (20) does not require technical rooms with dedicated climate control or precision HVAC infrastructure typical of conventional data centers.

[0253] Consequently, the equipment (20) can be installed in existing spaces such as parking lots, utility rooms, warehouses or technical areas, provided that sufficient general ventilation is provided to evacuate the hot air from the radiators, reducing investment in civil works and allowing the rapid deployment of computing capacity close to the points of consumption.

[0254] Orchestration of multiple devices (20) as a distributed microdata center

[0255] In a distributed mode, a plurality of containment equipment is interconnected through an orchestration layer defined by processor-executed instructions that consolidates the telemetry of each unit, including at least component temperatures, fluid flow rates, power consumption, processing load, and network link quality parameters.

[0256] The orchestration layer comprises a task scheduler and a communication optimization module configured to dynamically assign inference and / or training computing tasks among the plurality units, using low-communication distributed training algorithms that allow operation over moderate bandwidth networks.

[0257] In this way, the set of units operates as a distributed micro data center that simultaneously optimizes: (i) the use of computing resources, (ii) energy consumption profiles at the industrial plant or geographical area level, and (iii) the location of workloads with respect to their data sources, reducing dependence on large-scale centralized data centers.

[0258] In optional implementations, the orchestration layer is also configured to measure and record the use of computing and energy resources per unit, allowing the application of billing or compensation schemes between different organizations that share the distributed infrastructure.

[0259] Physical and operational safety specific to dielectric fluid

[0260] In other desirable embodiments, the equipment (20) comprises fluid level sensors, leak detectors, and one or more quick-closing valve actuators associated with instructions stored and executed by a control processor.

[0261] When abnormal conditions are detected, including a sudden drop in fluid level, presence of fluid in containment trays, or tilting of the casing outside a predetermined range, the system executes automatic safety procedures, including controlled shutdown of processing units and hydraulic isolation of the tank.

[0262] In a complementary mode, the equipment casing incorporates at least one biometric authentication device connected to the control processor, configured to authorize or deny the opening of hardware access panels based on biometric credentials of previously registered users, recording physical access events for auditing and traceability purposes. High-speed audio and video processing

[0263] In this configuration, the system receives, processes and correlates in real time video, still image and audio streams from different capture devices distributed in geographical spaces, such as fixed IP cameras, mobile cameras, thermal cameras, vehicle cameras, cameras mounted on drones and smart glasses worn by operators.

[0264] The advantages achieved in terms of audio and video processing include: reduced bandwidth required for telecommunications networks and the cloud, since instead of continuously transmitting high-resolution video streams to a remote data center, the device (20) can process these streams locally and output only events, descriptions, summaries, or compact metadata; operational resilience to network failures, as it can maintain AI processing locally, allowing the device (20) to continue analyzing video, audio, and process data even when external links are degraded or interrupted; improved protection of sensitive data, as stream processing takes place within the controlled environment where the device (20) is installed; and horizontal scalability through modular deployment, allowing the processing load to be distributed among different edge units.and energy optimization through selective local processing of relevant video and audio streams, allowing dynamic adjustment of GPU module activation and cooling. Technical effects

[0265] The modular and scalable distributed AI processing system (30) overcomes limitations of existing solutions by:

[0266] -Modular and Scalable Architecture: Attachable clamping structures allow for flexible configuration from individual units to massive deployments, with the addition / removal of components without affecting overall operation. Compact design with handles facilitates transport and rapid deployment.

[0267] -Optimized Edge Processing: Local AI execution reduces end-to-end latency, maintains privacy of sensitive data, and ensures continuous operation in the event of network failures. It transmits only aggregated information, minimizing bandwidth usage to the cloud.

[0268] -Advanced Cooling: Biodegradable dielectric media and ambient air eliminate dependence on water. AI controls dynamically adjust pumps, fans, and forced convection. Meter (103) optimizes energy consumption.

[0269] -Integrated Security: Thermal imaging cameras (200) predict fires using AI; suppression system (300) with self-directed nozzles (500); temperature sensors (101); biometric locks (105) with full traceability.

[0270] -Simplified Maintenance: Plug-and-play design with quick-connect interfaces, self-configuration, and labeled wiring. Double anchoring, a tilted tank with a drain valve, telescopic rails, and removable handles facilitate operation. Remote monitoring reduces on-site interventions.

[0271] -Operational Flexibility: Implements LLMs and processes audio, video, and image. Triple modularity allows for unit upgrades (GPU, NPU, ASIC, FPGA) without redesign. Indoor / outdoor installation with water resistance.

[0272] -Industrial Integration (OT): Direct connection with sensors, PLCs, and actuators. It executes digital twins (PINN), acting as an energy autopilot that optimizes overall consumption and extends the lifespan of GPUs.

[0273] -Infrastructural Efficiency: Airtight containment and external radiators eliminate the need for dedicated HVAC. Deployment in non-IT spaces (parking lots, industrial areas) without traditional data center infrastructure. Multi-device orchestration as a distributed microdata center reduces dependence on centralized centers.

[0274] -Operational Safety: Level / leak sensors, quick-closing actuators, tilt / abnormal movement detection and biometric authentication ensure system and environmental protection. Configurations and scalability

[0275] The system (30) can incorporate two to twenty high-performance portable modular AI processing systems (10), each with one to five uninterruptible power supply (UPS) containment systems (204) and one to five telecommunications unit (203) containment systems. Environmental monitoring is provided by two to four temperature sensors (101) and two to four humidity sensors (102). Controlled ventilation utilizes eight to eighty grille modules (104), scaling according to the installed systems. Security integrates two to four surveillance cameras (100), two to ten thermal imaging cameras (200), and one to four fire suppression systems (300) with integrated nozzle extinguishing systems (500). The mounting structures accommodate up to two processing systems per structure, allowing for expandable configurations via modular coupling. Examples

[0276] The following describes specific configurations of the modular and scalable distributed AI processing system (30), illustrating preferred implementation methods without limiting the scope of the invention as defined in the claims.

[0277] Example 1: Compact Configuration for Edge Computing Applications

[0278] In this modality, the system (30) comprises two portable, high-performance AI processing modular systems (10), each with the following specifications:

[0279] Each portable modular AI processing unit (20) incorporates a stainless steel tank measuring 90 centimeters long, 45 centimeters high, and 38 centimeters wide, along with a removable stainless steel lid measuring 90 centimeters long, 8 centimeters high, and 38 centimeters wide. The tank contains 100 liters of biodegradable dielectric oil as a cooling medium. The computing system includes an ATX motherboard, a 16-core CPU, 64 gigabytes of RAM, and 2 terabytes of NVMe storage.

[0280] Four GPUs are installed for AI processing, each with attachable handles on top for easy removal. The cooling system comprises two pumps, an additional backup pump, eight fans, two rear-mounted radiators, an additional redundant radiator, five heatsinks integrated into the case and lid, a forced convection stirring system with a backup agitator, a protective grille, four oil temperature sensors, and an AI-based electronic control unit. Two dehumidifiers are located on top of the tank, along with an internal humidity sensor, a biometric lock, an emergency button, four waterproof Ethernet connectors, and two power supply connectors.

[0281] The uninterruptible power supply (UPS) containment system (204) incorporates a stainless steel tank measuring 90 cm long, 48 cm high, and 52 cm wide, with a removable lid also measuring 90 cm long, 8 cm high, and 52 cm wide. It contains 120 liters of biodegradable dielectric oil. Three 3 kV UPS units are installed, each with four handles for easy handling, along with a surge protection device (SPD), a circuit breaker, an automatic transfer switch (ATS), two telescopic multi-device anchors, two dehumidifiers, an emergency stop button, four Ethernet connectors, and two power supply connectors.

[0282] The containment system for the telecommunications unit (203) incorporates a stainless steel tank measuring 90 centimeters long, 48 centimeters high, and 22 centimeters wide, with a removable lid measuring 90 centimeters long, 8 centimeters high, and 22 centimeters wide. It contains 60 liters of biodegradable dielectric oil. Two 48-port switches, two routers, one modem, three telescopic multi-device anchors, one dehumidifier, one biometric lock, one emergency button, four Ethernet connectors, and two power supply connectors are installed.

[0283] The monitoring system includes two surveillance cameras (100), two temperature sensors (101), two humidity sensors (102), four thermal imaging cameras (200), an energy consumption meter (103), sixteen grid modules (104), an additional biometric lock (105) on the external structure, and a fire suppression system (300) with its corresponding extinguishing system with integrated nozzles (500). The mounting structures accommodate the two systems (10) with height adjustment, a two-axis circular bubble level, solid anchor holes for floors and walls, and a bottom tray for leak detection. The complete system is installed in approximately two hours.

[0284] Example 2: High Capacity Configuration for Distributed Edge Data Center

[0285] In this high-capacity mode, the system (30) comprises ten portable modular high-performance AI processing systems (10), distributed in an industrial facility for decentralized artificial intelligence processing.

[0286] Each portable modular AI processing unit (20) maintains the same tank dimensions specified in Example 1 (ninety by forty-five by thirty-eight centimeters), but contains 180 liters of biodegradable dielectric oil for increased heat dissipation capacity. The computing system includes an E-ATX motherboard, a thirty-two-core CPU, two hundred and fifty-six gigabytes of RAM, and eight terabytes of NVMe storage. Eight state-of-the-art GPUs are installed, each with attachable handles.

[0287] The cooling system incorporates four pumps with two additional backup pumps, sixteen fans, four radiators (two rear and two front) with two additional radiators for redundancy, ten heatsinks integrated into the casing and lid, two forced convection stirring systems with two backup stirrers, two protective grilles, four strategically placed oil temperature sensors, and two AI-based electronic controls operating in redundant configuration. Also included are four dehumidifiers, two internal humidity sensors, a biometric lock, an emergency button, eight waterproof Ethernet connectors, and four power supply connectors.

[0288] The system incorporates three uninterruptible power supply (UPS) containment systems (204), each with the dimensions specified in Example 1 (90 by 48 by 52 centimeters) and containing 150 liters of biodegradable dielectric oil. Each UPS system houses six 5-kilovolt-ampere UPS units, with eight handling handles, two surge protection devices (SPDs), two circuit breakers, two automatic transfer switches (ATSs), five telescopic multi-device anchors, four dehumidifiers, an emergency stop button, eight Ethernet connectors, and four power supply connectors.

[0289] The system incorporates two telecommunications unit containment systems (203) with dimensions specified in Example 1 (ninety by forty-eight by twenty-two centimeters) and each containing 80 liters of biodegradable dielectric oil. Each system houses four 96-port switches, four high-capacity routers, two modems, six telescopic multi-device anchors, two dehumidifiers, a biometric lock, an emergency button, eight Ethernet connectors, and four power supply connectors.

[0290] The expanded monitoring system includes four surveillance cameras (100), four temperature sensors (101), four humidity sensors (102), ten thermal imaging cameras (200) strategically distributed to cover all systems (10), an energy consumption meter (103), eighty grille modules (104) providing scalable ventilation, four biometric locks (105) on external structures, and four fire suppression systems (300) with their corresponding extinguishing systems with integrated nozzles (500) distributed to provide redundant coverage of the entire facility.

[0291] The modular mounting structures connect to form an integrated infrastructure that houses the ten (10) systems. Each structure is equipped with a double-anchoring mechanism, variable height adjustment, a two-axis circular bubble level, solid anchor points, and lower drip trays. The entire system is interconnected by high-speed fiber optic cabling, distributed grounding devices, and integrated lighting to facilitate maintenance. A 100-kilovolt-ampere generator provides complete backup power for the entire installation.

[0292] Example 3: Intermediate Configuration for Hospital Applications

[0293] In this optimized mode for medical image processing and patient data management, the system (30) comprises five portable modular high-performance AI processing systems (10).

[0294] Each portable modular AI processing unit (20), with tank dimensions of 90 by 45 by 38 centimeters, contains 140 liters of biodegradable dielectric oil. The computing system includes a microATX motherboard, a 24-core CPU, 128 gigabytes of RAM, and 4 terabytes of NVMe storage. Six GPUs specialized in medical image processing, with attachable handles, are installed.

[0295] The cooling system comprises three pumps with an additional backup pump, twelve fans, three radiators (two rear and one front) with an additional redundant radiator, seven integrated heat sinks, a forced convection stirring system with a backup agitator, a protective grille, four oil temperature sensors, and an AI-based electronic control system. It also includes three dehumidifiers, two internal humidity sensors, a biometric lock with traceability logging for regulatory compliance, an emergency stop button, six waterproof Ethernet connectors, and three power supply connectors.

[0296] The system incorporates two uninterruptible power supply (UPS) containment systems (204) measuring 90 x 48 x 52 centimeters, each containing 130 liters of biodegradable dielectric oil. Each UPS system houses five 4kV UPS units, with six handles, two surge protection devices (SPDs), one circuit breaker, two automatic transfer switches (ATSs), three telescopic multi-device anchors, three dehumidifiers, an emergency stop button, six Ethernet connectors, and three power supply connectors.

[0297] The system incorporates two telecommunications unit containment systems (203) measuring 90 x 48 x 22 centimeters and containing 70 liters of biodegradable dielectric oil. Each system houses three 48-port switches, three routers, one modem, four telescopic multi-device anchors, two dehumidifiers, a biometric lock, an emergency button, six Ethernet connectors, and three power supply connectors.

[0298] The monitoring system includes three surveillance cameras (100), three temperature sensors (101), three humidity sensors (102), six thermal imaging cameras (200), an energy consumption meter (103), forty grid modules (104), three biometric locks (105) on external structures, and two fire suppression systems (300) with integrated nozzle extinguishing systems (500). Modular mounting structures house the five systems (10) with all the adjustment, leveling, and leak detection features described in previous examples. The system is interconnected via fiber optic cabling, includes grounding devices, integrated lighting, and a fifty-kilovolt-ampere generator for backup power. The complete installation ensures compliance with medical data privacy regulations by keeping all processing within the hospital premises.

[0299] These examples illustrate the flexibility and scalability of the system (30), allowing configurations adapted to different operational requirements without departing from the scope of the invention defined in the claims.

[0300] The modular and scalable distributed AI processing system (30) is suitable for industrial application in multiple sectors that require decentralized artificial intelligence processing with high energy efficiency and the ability to operate in diverse environments.

[0301] The invention can be manufactured and used in the distributed data center industry, edge computing infrastructure, AI-powered video and image processing facilities, and intelligent industrial monitoring systems. Its modular design allows for the mass production of the systems (10) and equipment (20), facilitating their commercialization and scalable deployment.

[0302] The system is applicable in manufacturing, mining, and oil and gas for local processing with operational resilience. In hospitals, it ensures the processing of medical images and patient data, keeping information within the facility. It is applicable in shopping malls and hotels for intelligent infrastructure management. The compact design allows for deployment in building parking garages, optimizing space, and in communication towers, providing local support for critical networks. In financial environments, the system (10) guarantees privacy and security by processing confidential data entirely locally, ensuring that sensitive financial information does not leave the premises, avoiding the risks of transferring data to remote servers or the cloud, and complying with strict data protection regulations.

[0303] The system's portability, environmental resistance and rapid installation capability (10) enable its deployment in remote locations, temporary facilities, military and emergency operations, as well as in critical infrastructures requiring high availability and operational continuity.

[0304] The invention can be industrially produced using standard manufacturing processes for electronic equipment, modular metal structures, and specialized cooling systems, ensuring its commercial viability and reproducibility on an industrial scale.

[0305] The foregoing description of the preferred embodiment enables persons skilled in the art to implement or use the present invention. For those skilled in the art, various modifications of these embodiments will be readily apparent, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the invention as defined in the claims.

[0306] Therefore, the present invention shall not be limited to the embodiments shown in the description, but shall be given the widest scope consistent with the novel principles and features described.

[0307] Although embodiments of the present invention have been described above with reference to the accompanying drawings, a person skilled in the art will understand that the present invention may be modified in various ways without departing from the scope and essential aspects defined in the claims. For example, those skilled in the art may modify a material or the size of each component according to a field of application, or may combine embodiments in ways not specifically described. However, such variations will not depart from the scope of the present invention. Therefore, it should be understood that the embodiment described above is merely illustrative in all respects, but in no way limits the scope defined in the claims.

Claims

A modular and scalable distributed AI processing system (30), comprising: at least one high-performance portable modular AI processing system (10) incorporating at least one portable modular AI processing unit (20), wherein the at least one unit (20) comprises a processing subsystem configured to process data using artificial intelligence, a biodegradable dielectric fluid immersion cooling subsystem, and an artificial intelligence and digital twin subsystem; a control and monitoring subsystem configured to supervise and control the at least one system (10); a power supply subsystem configured to supply power to the at least one system (10) and to the control and monitoring subsystem; and a telecommunications subsystem configured to enable communication between said subsystems;where the control and monitoring subsystem is configured to interconnect with a control and monitoring center (40) to transmit system status information.; The system (30) according to claim 1, wherein the processing subsystem incorporates an external protection and security subsystem and an internal protection and security subsystem. The system (30) according to claim 1, wherein the control and monitoring subsystem incorporates at least one server with an application for modular IA system management (600) that is interconnected with structured database servers (601), unstructured database servers (602), information backup servers (603) and the at least one system (10), through the telecommunications subsystem by means of wired and wireless communication protocols. The system (30) according to claim 3, wherein the server with application for modular IA system management (600) incorporates a management, control and monitoring platform (60) configured to manage and control the flow of information and the operation of the system devices. The system (30) according to claim 1, wherein the telecommunications subsystem operates by means of a telecommunications network (604) and Ethernet (614) with wired or wireless protocols. The system (30) according to claim 5, wherein the telecommunications subsystem incorporates containment systems for a telecommunications unit (203) and allows one or multiple devices to be connected to the system (30) through a wired network infrastructure based on industrial Ethernet, by means of one or more PoE (Power over Ethernet) capable switches that provide power and IP connectivity. The system (30) according to claim 5, wherein the connection is established wirelessly using means selected from the group: industrial Wi-Fi access points, dedicated radio links, cellular links (4G / 5G / 6G) and satellite links, in which case the field devices transmit data streams to one or more Video and Audio Gateway systems (611) which forward said streams to the system (30) via Ethernet links or optical links. The system (30) according to claim 1, wherein the at least one piece of equipment (20) is interconnected with external devices selected from the group: IoT devices (605), desktop user stations (606), portable user stations (607), mobile communication devices (608), mobile communication and control devices (609), OT devices (610) and video and audio Gateway systems (611). The system (30) according to claim 8, wherein the IoT devices (605) are one or more selected from the group: smart glasses, smartwatches, vibration sensors, temperature sensors, pressure sensors, acoustic sensors, RFID tags, Bluetooth beacons, smart readers, programmable logic controllers (PLCs), connected robots, air quality sensors, water leak sensors, smart thermostats, smart lighting systems, occupancy sensors, smart security cameras, surveillance systems, electronic access control systems, and edge computing devices. The system (30) according to claim 2, wherein the external protection and security subsystem incorporates external support structures (107) to which are interconnected: surveillance cameras (100), temperature sensors (101), humidity sensors (102), energy consumption meters (103), grid modules (104) and biometric locks (105). The system (30) according to claim 10, wherein the grid module system (104) is configurable, modular and expandable according to the number of containment systems for high-performance processing units (201), and allows the entry of ambient air for cooling. The system (30) according to claim 2, wherein the internal protection and safety subsystem comprises internal support structures (106) to which are interconnected: thermal imaging cameras (200), containment systems for high-performance processing units (201), support structures for housing processing units (202), containment systems for a telecommunication unit (203), containment systems for an uninterruptible power supply unit (204), trays for capturing dielectric fluid leaks (205), an electric generating motor (206), a floor anchoring system with adjustable height (207), fire suppression systems (300), support structures with a double anchoring mechanism (400), and fire extinguishing systems with integrated nozzles (500). The system (30) according to claim 12, wherein the thermal imaging cameras (200) integrate artificial intelligence algorithms configured to predict fire risks by predictive analysis, and the fire extinguishing system with integrated nozzles (500) is configured for automatic orientation towards the fire generation point detected by the thermal imaging cameras (200). The system (30) according to claim 12, wherein the internal support structures (106) are equipped with a double anchoring mechanism (400) comprising a shaft and a stabilizing arm, configured to position the portable modular AI processing equipment (20) in an ergonomic position for maintenance operations. The system (30) according to claims 10 and 12, wherein the internal support structures (106) and the external support structures (107) are modular, assembleable, disassemblable and coupleable to each other, with anchors designed to organize different types of cables selected from the group: fiber optic, Ethernet and power cables. The system (30) according to claim 12, wherein the internal support structures (106) allow the height to be adjusted to separate or bring closer the different containment systems for high-performance processing units (201), have a circular bubble level for two axes, and have solid anchor holes for floor and walls. The system (30) according to claim 12, wherein the internal support structures (106) at the bottom have a tray (205) designed to contain possible leaks of dielectric fluid from the portable modular AI processing equipment (20). The system (30) according to claim 12, wherein the containment systems for high-performance processing units (201) are designed with an incline on the tank floor which, together with a valve, allows the implementation of a draining mechanism to facilitate maintenance. The system (30) according to claim 1, wherein the system control and monitoring center (40) is interconnected with the servers with application for modular IA system management (600), and through a telecommunications network (604) receives alert and notification information from all elements, parts and devices of the system (30). The system (30) according to claim 1, wherein the power supply subsystem incorporates an electric generating motor (206) interconnected with each of the following components: portable modular AI processing equipment (20), surveillance cameras (100), temperature sensors (101), humidity sensors (102), energy consumption meter (103), biometric locks (105), thermal imaging cameras (200), containment systems for telecommunication unit (203), containment systems for uninterruptible power supply unit (204), fire suppression systems (300) and fire extinguishing systems with integrated nozzles (500). The system (30) according to claim 1, characterized by triple modularity comprising: hardware modularity by means of interchangeable accelerated processing units in at least one device (20) selected from the group: GPU, NPU, AI ASIC, FPGA; physical format modularity by means of scalable containment systems; and software modularity by means of an orchestration layer with processor-executed instructions for dynamic allocation of workloads among multiple devices (20). The system (30) according to claim 1, wherein the at least one piece of equipment (20) incorporates a language model platform (50) corresponding to processor-executed instructions with generative AI algorithms. A portable modular AI processing equipment (20), comprising: a modular structure and containment subsystem defining an interior housing volume; a processing subsystem housed within the interior volume, including at least one accelerated processing unit and computing hardware (704); an advanced cooling subsystem comprising a biodegradable dielectric fluid (705) in direct contact with the processing subsystem components, and a heat dissipation system configured to transfer heat from the dielectric fluid to the outside of the interior volume; and an artificial intelligence and digital twin subsystem configured to run physics-informed neural network (PINN)-based digital twin models; wherein the equipment (20) is configured for operation free of dedicated external climate control infrastructure. The portable modular equipment (20) according to claim 23, further comprising: a power supply subsystem configured to supply power to the equipment components; an OT / IoT industrial communications and integration subsystem configured for data exchange with external devices; a monitoring and control subsystem configured to monitor operating parameters of the equipment; a distributed orchestration and management subsystem; and a real-time multimedia processing subsystem. The portable modular equipment (20) according to claim 23, wherein the containment and modular structure subsystem comprises: a base box (22) comprising a container tank (709) that defines the internal housing volume; adjustable rails housed for dynamic reconfiguration of internal components; a top cover (21) assembled to the base box (22); and modular anchoring and interconnection mechanisms interconnected externally to the base box (22); said subsystem having an incline on the floor of the tank (709) and a valve for emptying and maintenance. The portable modular equipment (20) according to claim 25, wherein the container tank (709) is made of stainless steel. The portable modular equipment (20) according to claim 25, wherein the modular anchoring and interconnection mechanisms comprise: an anchoring mechanism with a structure (710) configured to allow coupling with other modular equipment (20) and gripping handles (707) for transporting the equipment. The portable modular equipment (20) according to claim 23, wherein the processing subsystem comprises: a GPU cabinet (701) housing at least one interchangeable accelerated processing unit; the CPU (734) for instruction processing; and the computing hardware (704) including: a board (714) for system control, a motherboard (724) for interconnection and control of electronic components, RAM memory (744) for volatile data storage, and a storage system (754) for non-volatile storage of data and information; a communications system (764) for interconnecting the equipment (20) with other apparatus and devices for sending and exchanging information; and sensors (814). The portable modular equipment (20) according to claim 28, wherein the GPU cabinet (701) allows for multiple accordion-style parallel GPU configurations, controlling different physical separations between each of the GPUs by means of instructions executed by a processor, for optimization of cooling and temperature. The portable modular equipment (20) according to claim 28, wherein the communication system (764) comprises industrial communication modules configured to interface with Operational Technology (OT) systems by means of wired protocols selected from the group: Modbus TCP, PROFINET, EtherNet / IP, RS-485 / 422, Industrial Ethernet, OPC-UA; and wireless protocols selected from the group: Wi-Fi, Bluetooth / BLE, LTE / 5G, NB-IoT, LoRaWAN, WirelessHART, ISA100.11a. The portable modular equipment (20) according to claim 30, wherein the communication system (764) is configured to: receive real-time data from selected external industrial devices of the group: industrial sensors, programmable logic controllers (PLCs) and actuators; generate control signals intended to adjust operating parameters of coupled industrial equipment, based on the processing of the received data by means of artificial intelligence algorithms executed in the processing subsystem; and receive and process in real time multimodal data streams from selected capture devices of the group: IP interface video surveillance cameras, motorized PTZ cameras, thermal cameras, cameras mounted on unmanned vehicles, wearable body cameras, and capture devices integrated into smart glasses with microphones for audio. The portable modular equipment (20) according to claim 23, wherein the cooling subsystem comprises: a coolant and insulator (705) consisting of biodegradable dielectric fluid contained within the inner volume of the base case (22) and in direct contact with the at least one accelerated processing unit and computing hardware (704); a heat dissipation system comprising: at least one pump configured to circulate the dielectric fluid, at least one radiator (813) configured to transfer heat from the dielectric fluid to the outside, at least one fan (810) associated with the radiator (813), and cooling system containment units (811);AI-adaptive electronic controls (812) configured to dynamically regulate the operation of the heat dissipation system, to detect pump and radiator failures, and to redirect the flow of the dielectric fluid using redundant pumps and radiators by means of a solenoid-based locking system; wherein the coolant and insulator (705) absorbs the heat generated by the processing components. The portable modular equipment (20) according to claim 32, comprising at least one heat sink (706) interconnected externally to the base case (22), distributed around the base case (22) for passive thermal dissipation. The portable modular equipment (20) according to claim 32, wherein the AI ​​adaptive electronic controls (812) are configured to dynamically adjust selected group parameters: pump flow rate, fan speed (810), forced convection system RPM, and physical separation between processing units by means of adjustable rails. The portable modular unit (20) according to claim 23, wherein the top cover (21) houses dehumidifiers (703) configured to remove internal vapors and protect the electronic components from moisture. The portable modular equipment (20) according to claim 24, wherein the power supply subsystem comprises: power supplies (702) configured to supply electrical power to the equipment components; an uninterruptible power supply (UPS) system configured to ensure operational continuity in the event of failures in the external power network; wherein the power supplies (702) are interconnected with the processing subsystem, the cooling subsystem, and the monitoring and control subsystem. The portable modular equipment (20) according to claim 24, wherein the monitoring and control subsystem comprises: a display with a biometric authentication system (800) configured to control access to the equipment by means of identity verification; an AI display for temperature regulation (801) configured to monitor system conditions and adjust operating parameters to maximize efficiency and safety; sensors (814) configured to measure operating parameters of the equipment; wherein the display with a biometric authentication system (800) and the AI ​​display for temperature regulation (801) are interconnected to the top cover (21), and the sensors (814) are interconnected with the computing hardware (704). The portable modular equipment (20) according to claim 37, wherein the screen with biometric authentication system (800) is interconnected with one or more biometric elements selected from the group: fingerprint recognition sensors, iris recognition sensors, voice tone recognition sensors, facial recognition sensors, to identify the veracity of the identity of people. The portable modular equipment (20) according to claim 37, wherein the sensors (814) are interconnected with the computing hardware (704) and are from the group that measures selected physical variables from: temperature, pressure, force, position, displacement, light, magnetic fields, gases, humidity, and fluid level. The portable modular equipment (20) according to claim 37, wherein the AI ​​temperature regulation display (801) is configured to receive data from the sensors (814) and send instructions to the AI ​​adaptive electronic controls (812) of the cooling subsystem to dynamically adjust selected group parameters: pump flow rate, fan speed (810), and physical separation between processing units. The portable modular equipment (20) according to claim 25, comprising: a drainage system (708) interconnected to the base box (22), configured to facilitate the evacuation of the dielectric fluid during maintenance operations; said drainage system (708) comprising an incline in the floor of the base box (22) and a valve configured to facilitate the complete emptying of the dielectric fluid; card handles (700) configured to facilitate the removal and insertion of internal components of the processing subsystem; telescopic rails for vertical extraction of base plates and components; and a double anchoring system for stable positioning. The portable modular equipment (20) according to claim 23, wherein the artificial intelligence and digital twin subsystem comprises processor-executed instructions configured to execute at least one physically informed neural network (PINN)-based digital twin model of an external physical process, generating control signals to optimize the energy consumption of the external physical process and the equipment (20) itself. The portable modular equipment (20) according to claim 42, wherein the artificial intelligence and digital twin subsystem is configured to run at least one large language model (LLM) that communicates with the PINN models of the digital twin to interpret their outputs and produce structured control instructions or recommendations in natural language. The portable modular equipment (20) according to claim 42, wherein the control signals comprise selected elements from the group: setpoints for dielectric fluid recirculation pump speed, setpoints for radiator fan speed, hydraulic circuit valve settings, and commands to actuators that modify the relative position of GPU modules by means of adjustable rails to optimize dielectric fluid distribution and internal temperature. The portable modular equipment (20) according to claim 24, wherein the real-time multimedia processing subsystem comprises: multimedia capture interfaces that receive data from selected devices from the group: CCTV-type video surveillance cameras with IP interface, motorized PTZ cameras, thermal or infrared cameras, cameras mounted on unmanned aerial vehicles (drones), portable bodycam-type cameras, and capture devices integrated into smart glasses or smart helmets with microphones for audio; stream processors that execute multimodal correlation algorithms in the processing subsystem; and compression systems that generate compact events, descriptions, summaries, or metadata, reducing the bandwidth required to external networks. The portable modular equipment (20) according to claim 24, wherein the orchestration and distributed management subsystem comprises: a task scheduler configured to dynamically allocate inference and training workloads among multiple pieces of equipment (20); a communication optimization module that utilizes low-communication distributed training algorithms; a consolidated telemetry system configured to collect temperature, flow rates, power consumption, and link quality data from each piece of equipment (20); wherein the orchestration subsystem is configured to simultaneously optimize: utilization of computing resources, energy consumption profiles, and load location with respect to data sources.

Images