The Ethon Industrial AI Platform is a software platform for monitoring, analyzing, optimizing, and controlling manufacturing processes using artificial intelligence. It is offered as a Software as a Service (SaaS) and integrates data ingestion, contextual process modeling, and workflow applications into a single environment. It enables production teams, process engineers, and quality assurance staff to identify improvement opportunities, maintain process stability, and ensure full traceability of production data.

In addition to the Industrial AI Platform, Ethon offers a Quality Inspection add-on. Parts of this add-on can be deployed on edge devices.

Core functionality in the Industrial AI Platform

Data contextualization: the Platform integrates diverse manufacturing data sources,  including sensor readings, contextual process information (e.g., line stops, setpoints, machine status codes, product variants, and defect types), and historical records, into a unified environment.

Industrial AI models, algorithms, and agents: these provide the underlying capabilities spanning process observability, anomaly detection, root cause analysis, setpoint optimization, assisted and automated shopfloor decision making, and operational rule monitoring in a unified environment.

Configuration and setup, such as user management, status overviews of data source, process modeling, the definition of customer specific production event types, etc. Some of these configurations adapt Ethon's Industrial AI models to the specifics of a given customer's production setups.. 

Workflows are built atop the core functionality. They support consistent and intelligent production across shifts by providing timely information, recommended actions, and closed-loop control capabilities.

The Root Cause Analysis workflow is built to quickly and reliably uncover the causes of production issues. The platform aggregates all relevant parameters using the process knowledge graph, and its Reasoning Model builds a causal graph that reflects the physical cause-and-effect relationships within the production process. Once sufficiently integrated with production data, it continuously scans process parameters to detect anomalies and emerging issues in real time. The resulting incident reports are pre-populated with causal diagnostics. Root cause analysis can also be run ad-hoc, for example when process engineers need to bring up new lines or deep dive into historical data. Whether automatic or manually run, analysis results are presented first as intuitive summary descriptions with compelling visuals, and then allow interactive exploration to thoroughly understand the findings. This is supported via causal graphs, data statistics, and flexible rich data visualization. Production teams use these to understand the drivers of quality issues and process losses, across process settings and variables.

The Process Optimization workflow is built to continuously keep your production running optimally. It monitors process parameters, machine setpoints, and operational outcomes in real time. It identifies suboptimalities and recommends improvements. Configurable dashboards allow users to visualize process performance across lines, shifts, and batches. Users can also explore parameter trends and KPIs at multiple levels of granularity. Agents detect inefficiencies and emerging patterns using causal AI models. These models are grounded in a knowledge graph that reflects the physical structure of the production line. The agents compute optimal setpoints for quality and throughput. If needed, the underlying capabilities can also be run manually, for example by process engineers to solve constrained optimization involving multiple setpoints at the same time. This can for example be useful to virtually trial improvement actions before implementation, and significantly reduce the need for physical trials.

The Process Control workflow is built to keep production stable under changing conditions, across shifts and production lines. It monitors critical process parameters in real time to assess deviations from defined operating conditions. The causal AI models how effects propagate through the process, and thus identifies the most effective ways to counter drift. The workflow automatically recommends adjustments, and can be set up to directly apply them within safe operating bounds defined by engineers. Configurable rule monitoring, threshold-based alerting, and closed-loop causal steering are the key modalities of this workflow. In addition, live shopfloor dashboards show real-time and historical views of process performance. The dashboards are built from a powerful set of visualizations, including run charts, status indicators, and parameter overviews. Dashboards can be fully customized to fit the needs of your operators, process engineers, and shift leaders.

The Early Fault Detection workflow is built to proactively recommend actions before processes deviate, or equipment fails. This enables teams to address developing issues before they escalate into unplanned line stops, or scrap events. To set up early fault detection, users define the set of parameters to monitor and give examples of healthy behavior to the system, simply by selecting time ranges or batches that were particularly good. The platform then learns the healthy baseline across all monitored parameters, using a combination of causal AI and multivariate anomaly detection. This step is quick and easy, the user does not need to provide finegrained thresholds or similar manual input. Once the baseline is trained, this workflow continuously monitors production data to identify anomalies in process behavior or equipment conditions. It flags deviations in real time, and suggests corrective actions based on the current process state. In addition, users can explore the trends and deviations in the parameters using rich and configurable visualizations at the level of individual variables or entire production lines.

The Production Traceability workflow allows you to reconstruct the true flow of every part and batch throughout production. When you have event logs for each unit, the platform can dynamically apply process mining and visualize material movements, work-in-progress, throughput, and bottlenecks. For example, you can quickly identify unintended loops and rework. The platform also performs Value Stream Mapping based on routing and timing data. This allows users to identify and quantify value-adding activities, idle time, and how bottlenecks shift. To round out the traceability, the platform maintains searchable, permanent records of each product unit's history, including timestamps, routing, inspections, measurements, and image metadata. This enables use cases such as operational reviews, customer reporting, and audits.

In addition to these workflows, the platform also includes a data science environment giving experts direct access to the underlying cleaned and contextualized production data. This eliminates the need to rely on fragmented external files, and provides a single source of truth. The environment is built atop the popular JupyterHub system, such that users can build custom analyses and visualizations directly with Python notebooks.

The Ethon Quality Inspection add-on is built to collect inline quality measurements using a highly scalable computer vision software. It is compatible with any industrial camera and can for example detect and count defects. The add-on then connects these inspection data with the process parameters in the platform, enabling integration with the Root Cause Analysis workflow to identify the factors driving quality outcomes.

The add-on is available in two versions:

Inspector 1.0 is available as an on-premise solution and includes the functionality previously provided by the “Inspector” module. AI models are trained on sets of defect-free product images and measure deviations from the established norm. The workflow integrates with vision systems compatible with any industrial camera. It comprises three components:

1. The Inspector Backoffice has a web-based user interface to train, evaluate, and maintain/monitor inspection jobs. AI models trained with the Inspector Backoffice receive an image as input and produce a binary inspection result (i.e., OK, NOK). It can optionally receive and emit other information, for example provide explanations for the AI output, such as heatmaps showing where in the image a defect is suspected; receive product identifiers directly, or use an additional AI model to infer them automatically; and AI models may also read barcodes in various formats, as well as text, via optical character recognition.
2. The Inspector Lite can deploy the AI models from the Inspector Backoffice on the edge, close to the production line. The Inspector Lite can be installed on every inspection station, which allows for distributed computing within the factory. This may naturally reduce latency, and facilitate horizontal scaling of inspection capabilities within the factory.
3. The Factory Frontend is an interactive graphical user interface for sending new inspections to the Inspector Backoffice and visualizing the inspection results, connecting directly with up to four cameras.

Inspector 2.0 is a cloud-native evolution of Inspector 1.0 that modernizes the system architecture and enables the efficient use of deep learning models. Training, model management, and data processing are performed in the cloud, while inference runs on edge devices close to the production line to ensure low latency and reliable operation.

The system is powered by a vision foundation model trained specifically for manufacturing environments. Customers can adapt this model to their specific inspection tasks by providing a small number of reference images or annotations. These data are used to fine-tune the foundation model in the cloud. The resulting model can then be deployed to one or more edge inspection stations. This workflow remains hardware-agnostic and compatible with any industrial camera.

This architecture reduces the amount of labeled data required to configure inspections while enabling new inspection use cases beyond fixed layouts, such as conveyor-based inspection or flexible product configurations. Inspection results integrate directly with the Ethon Industrial AI Platform and can be used as structured quality signals in workflows such as root cause analysis and process monitoring.

Overview

The EthonAI Inspector is a software to detect quality defects using artificial intelligence (AI). The AI is trained on sets of defect-free product images, and measures deviations from the norm. The EthonAI Inspector is available as an on-premise solution.

Details

To accommodate the varied deployment challenges of modern manufacturing, the EthonAI Inspector consists of three distinct software components: the Inspector Backoffice, the Inspector Lite, and the Factory Frontend.

The Inspector Backoffice has a web-based user interface to train, evaluate, and maintain/monitor inspection jobs. AI models trained with the Inspector Backoffice receive an image as input and produce a binary inspection result (i.e., OK, NOK). It can optionally receive and emit other information, for example:

• provide explanations for the AI output, such as heatmaps showing where in the image a defect is suspected;
• receive product identifiers directly, or use an additional AI model to infer them automatically; and
• AI models may also read barcodes in various formats, as well as text, via optical character recognition.

The Inspector Lite can deploy the AI models from the Inspector Backoffice on the edge, close to the production line. The Inspector Lite can be installed on every inspection station, which allows for distributed computing within the factory. This may naturally reduce latency, and facilitate horizontal scaling of inspection capabilities within the factory.

The Factory Frontend is an interactive graphical user interface for sending new inspections to the Inspector Backoffice and visualizing the inspection results, connecting directly with up to four cameras.

Overview

The EthonAI Analyst is a software to monitor, analyze, and optimize manufacturing processes, combining the functionality of the previous EthonAI Analyst, EthonAI Tracker, and EthonAI Miner products. It enables process engineers, line managers, and quality teams to identify and act on the most impactful opportunities for improvement, maintain process stability, and ensure full traceability of production data. Users can access all relevant production information in one platform, obtain actionable insights, and explore data from the level of individual product units to factory-wide value streams.

Details

The Analyst uses artificial intelligence (AI), including proprietary Causal AI models and advanced process mining, to identify correlations and true cause-and-effect relationships in manufacturing data that are difficult to find using classical statistics. It supports monitoring, agentic workflows, analysis, simulation, material flow visualization, value stream mapping, product tracking, and custom analytics.

In the Monitoring workflows: production process parameters, key performance indicators (KPIs), and product history can be visualized and explored. Depending on the data integration pipeline, for example with MQTT, this can be achieved in near real-time. A configurable dashboard and fine-grained data exploration capabilities allow users to examine performance across lines, shifts, and batches.

The Analysis workflows use a Process Model to represent the ontology of a production line in detail. The integrated Causal AI identifies and displays cause-and-effect relationships between process parameters and outcomes such as yield, scrap rate, or product properties. Results are presented as interactive causal graphs with supporting data statistics.

The Simulation workflows provide a digital design of experiments to virtually test what-if scenarios and proposed process parameter changes before implementation. This reduces the need for physical trials, speeds up process optimization, and increases confidence in improvement actions.

Material Flow Analysis uses process mining on event log data to dynamically visualize material movements, work-in-progress, throughput, and bottlenecks, including unintended loops and rework. Value Stream Mapping automatically extracts routing and timing data to identify value-adding activities, idle time, and shifting bottlenecks, providing interactive exploration beyond static value stream snapshots.

The product tracking functionality provides searchable, permanent records of each product unit’s history, including timestamps, routing, inspections, measurements, and image metadata. Images are securely stored and archived in accessible formats, enabling operational reviews, customer reporting, and audits.

The Console, a built-in Jupyter-based environment, enables custom data analysis, giving experts direct access to cleaned and contextualized production data for bespoke analyses and visualizations without the need for fragmented files.

Overview

The EthonAI Operator is a software module for real-time monitoring, control, and optimization of manufacturing processes. The Operator builds on the Causal AI technology of the EthonAI Analyst to integrate real-time data, generate safe and optimized process parameters, and support automated decision-making on the shop floor. It combines process observability, anomaly detection, automated setpoint generation, and operational rule monitoring in a single environment.

The Operator is designed for use by production teams, process engineers, and quality assurance staff. It supports consistent and intelligent production across shifts by providing timely information, recommended actions, and closed-loop control capabilities.

Details

The Operator integrates diverse manufacturing data sources, including sensor readings, contextual process information (e.g., line stops, setpoints, machine status codes, product variants, and defect types), and historical records of issue resolution. Its core functions include:

The Process Observability feature continuously monitors production to support timely and well-informed decisions. Users can explore process parameters, trends, and deviations through configurable visualizations and real-time alerts. Data can be examined from the level of individual process variables to entire production lines, enabling early detection of anomalies and proactive intervention.

The Operator includes Rule Monitoring workflows for defining and monitoring specification limits for process parameters. It alerts users when parameters exceed thresholds and recommends corrective actions to maintain processes within defined tolerances.

The Operator provides customizable live Shopfloor Dashboards for real-time and historical monitoring of process performance. Dashboards can include visualizations such as run charts, status indicators, and parameter overviews. They can be tailored to specific user roles or operational needs and thus support both day-to-day operational oversight and long-term performance analysis.

The EthonAI Operator can be deployed as a standalone module or in combination with the EthonAI Analyst. Together, they provide integrated workflows for monitoring, analysis, and optimization of manufacturing processes.

Overview

The EthonAI Analyst is a software to monitor and analyze key metrics from production lines, for example to remedy quality issues or yield losses. Process engineers use the software to access all relevant data in one platform and get relevant insights about process parameters that are suspected to negatively affect production output.

Details

The Analyst uses an artificial intelligence (AI) to find correlations that are very challenging and onerous to find using classical statistics. The main workflows provided by the Analyst are tailored for process engineers and line managers who need to keep production running smoothly as well as gain deeper insights into the production lines, with the aim of improving them. These workflows are Monitoring, Analysis, and Simulation.

In the Monitoring workflows: production process parameters and KPIs can be visualized and explored. Depending on the data integration pipeline, for example with MQTT, this can be achieved in near real-time. In addition to fine-grained data exploration workflows, the monitoring capabilities include a customizable Dashboard.

The Analysis workflows are based on a Process Model that represents the ontology of a production line in a fine-grained manner. The analysis uses a “Causal AI” to find and present hidden correlations and related data statistics. Results are presented with an interactive display of graph-based relations and statistics from the historical data that underpins each analysis.

Using the Simulation workflows, based on the findings of a particular analysis, users can run virtual optimizations of setpoints and other parameters. These virtual what-if scenarios significantly speed up the task of planning and executing improvement actions on the factory floor.

Overview

The EthonAI Inspector is a software to detect quality defects using artificial intelligence (AI). The AI is trained on sets of defect-free product images, and measures deviations from the norm. The EthonAI Inspector is available as an on-premise solution.

Details

To accommodate the varied deployment challenges of modern manufacturing, the EthonAI Inspector consists of three distinct software components: the Inspector Backoffice, the Inspector Lite, and the Factory Frontend.

The Inspector Backoffice has a web-based user interface to train, evaluate, and maintain/monitor inspection jobs. AI models trained with the Inspector Backoffice receive an image as input and produce a binary inspection result (i.e., OK, NOK). It can optionally receive and emit other information, for example:

• provide explanations for the AI output, such as heatmaps showing where in the image a defect is suspected;
• receive product identifiers directly, or use an additional AI model to infer them automatically; and
• AI models may also read barcodes in various formats, as well as text, via optical character recognition.

The Inspector Lite can deploy the AI models from the Inspector Backoffice on the edge, close to the production line. The Inspector Lite can be installed on every inspection station, which allows for distributed computing within the factory. This may naturally reduce latency, and facilitate horizontal scaling of inspection capabilities within the factory.

The Factory Frontend is an interactive graphical user interface for sending new inspections to the Inspector Backoffice and visualizing the inspection results, connecting directly with up to four cameras.

Overview

The EthonAI Miner is a software to extract and analyze production history event logs to identify bottlenecks, process variations, and excess inventory. It helps manufacturers optimize throughput based on clearly visualized insights into their material flows.

Details

The Miner provides a comprehensive visual overview of material flow, based on advanced process mining techniques; to automate value stream mapping so it can be run dynamically with low overhead; and to monitor inefficiencies in real-time.

Overview

The EthonAI Observer is a software that uses artificial intelligence (AI) for real-time monitoring of manufacturing processes.Its advanced anomaly detection capabilities for multivariate process data help identifying unusual patterns or deviations across multiple correlated variables, signaling potential faults or issues in the production process.

Details

The key functionalities of the Observer revolve around integrating and managing diverse data sets from various manufacturing sources; real-time alerting based on discovered anomalies; as well as customized and detailed analytics on operational performance.

Overview

The EthonAI Tracker is a software to streamline the quality management and monitoring of physical products throughout their lifecycle. It offers detailed access to each product unit’s history and images in a single platform.

Details

The Tracker provides comprehensive tracking of individual product units. It captures critical details such as time, routing, inspections, measurements, and image metadata. It gives access to the full operational history of each product unit by searching serial, batch, or order numbers. Users can easily find out when and where each product unit was processed, and also under what specific conditions. Permanent data storage enables future retrieval for transparency.

The Tracker has subsystems for image storage and data management.

• Images are securely stored and archived on the cloud in accessible formats, supporting various operational and customer service activities.
• Direct sharing of reports with customers in case of damage (e.g., via a PDF export function)
• Automatic data compilation at the order level, for a complete overview of each batch.
• Systematic archiving of all data, including images, for permanent access and reference.

Overview

The EthonAI Console is a scripting environment for custom data analysis of manufacturing processes. Users can create and manage fully custom analytics and reports using the widespread Python programming language.

Details

The core functionalities of the Console include the ability to integrate and manage diverse data sets from various manufacturing sources; use the Python programming language, via Jupyter notebooks, to create custom analytics tailored to specific manufacturing needs; and generate comprehensive reports and insights for strategic decision-making and process optimization.