DX-AI IoT Prism
2025-05-10
An integrated AI framework for effectively processing, analyzing, and visualizing data in a manufacturing IoT environment
The "DX-AI IoT Prism" describes an integrated framework for effectively processing, analyzing, and visualizing the data collected in a manufacturing IoT environment. This system supports data-driven decision making and provides real-time monitoring and predictive analytics capabilities. Specifically, it helps non-experts easily leverage the value of IoT data through AI-based data analysis, automated analysis curation, and generative AI-powered question-answering. The system architecture consists of data collection, processing, storage, analysis, and visualization stages, and presents various visualization modules and future expandability. The core values are intelligent interaction, in-depth analysis support, efficient decision-making support, data integration and accessibility, and intuitive visualization.
Overview
DX-AI IoT Prism is an integrated framework for effectively processing, analyzing, and visualizing the diverse sensor data collected in an IoT (Internet of Things) environment. This system aims to support data-driven decision making and provide real-time monitoring and predictive analytics capabilities. It offers visualization components with a consistent "Prism" theme to ensure a seamless user experience.
Automated Analytics Curation
It automatically detects data patterns and recommends the optimal visualization methods, curating effective analysis directions to accelerate the user's discovery of insights.
AI-based Data Analysis
For each visualization analysis, it provides better insights through generative AI. This enables users to perform more accurate and faster analyses.
AI-based Analytical Q&A
Utilizing generative AI, users can ask questions about the data in natural language. The system understands the questions, analyzes the relevant data, and provides insightful answers, allowing even non-technical users to easily leverage the value of IoT data.
Data Visualization
It presents the analysis results in intuitive visualization formats, helping users easily understand the data and take quicker actions. Data visualization supports efficient decision-making and increases the transparency of business processes.
Data Analysis
It performs in-depth analysis based on the collected data, providing valuable business insights and extracting meaningful information from the data. By visually representing the data analysis results, it saves time and cost, and aids in decision-making.
Data Processing
It effectively processes the data collected from various IoT sensors. The processed data is presented in an intuitive visual form, helping users easily identify patterns and trends, and supporting effective decision-making.
System Architecture
The overall system architecture of the DX-AI IoT Prism consists of stages for data collection, processing, storage, analysis, and visualization.
Data Collection
Data is collected from various IoT sensors and devices.
Data Processing
The collected raw data is refined and processed into an analyzable format.
Data Storage
The processed data is efficiently stored and managed.
Data Analysis/Modeling
Statistical analysis, pattern recognition, and predictive modeling are performed based on the stored data.
Visualization and Dashboard
The analysis results are provided in various chart and dashboard formats for easy user understanding.
System Architecture
1. Time Series Analysis Module
This module provides the functionality to analyze the temporal changes in data and predict trends. The analyzed results can be visualized as time series graphs to identify data patterns over time. This allows you to understand the inherent patterns or periodicity in the data, which can provide useful information for predictive modeling.
timeseries.py
Visualizes the trend of data changes over time. It can include basic time series line charts and more.
trend.py
Analyzes and visualizes the long-term trend of data. It can utilize moving averages, regression lines, and other techniques.
pattern.py
Explores and visualizes patterns in the data that repeat or follow specific rules.
2. Distribution Analysis Module
You can visually analyze the distribution of values and confirm the distribution and outliers of the data through histograms and box plots. This will help you better understand the characteristics of the data.
histogram.py
Visualizes the frequency distribution of data in the form of a bar graph.
boxplot.py
Creates a box plot that visually represents the quartiles, median, and outliers of the data. It is useful for comparing the distribution between groups.
distribution.py
Provides visualizations to understand the distribution characteristics (mean, variance, skewness, kurtosis, etc.) of the data. It may include a histogram and a kernel density estimation (KDE) plot.
3. Pattern Analysis Module
It creates heatmaps or scatter plots to understand the correlation between data variables. It is used to visually analyze the interaction and patterns between the data.
heatmap.py
It creates a heatmap that visually represents the data matrix using colors to identify patterns or relationships.
correlation.py
It analyzes and visualizes the correlation between multiple variables. It can use heatmaps or scatter plot matrices.
clustering.py
It visualizes the results of clustering analysis, which groups similar data points together. It can display the cluster results on a scatter plot using colors or markers.
4. Advanced Analytics Module
By analyzing the frequency, outliers, and periodicity of data, we provide insights that can be applied in practice from the perspective of facility/equipment operation.
fft.py
Analyzes the frequency components of time series data using the Fast Fourier Transform and visualizes the spectrum. Can be used for vibration data analysis, among other applications.
timefreq.py
Visualizes the results of time-frequency analysis. May include spectrograms and other related visualizations.
anomaly.py
Detects and visualizes anomalous (outlier) data that falls outside the normal range. May include features such as highlighting points that exceed a certain threshold in time series data.
cycle_analysis.py
Analyzes the operation/shutdown cycles of facilities/equipment and detects abnormal cycles based on time and status value.
5. 3D, Alarm Analysis
threed.py
Visualizes data in a 3D space. Can create 3D scatter plots, surface plots, and more.
alarm_threshold.py
Provides visualization comparing set alarm thresholds with actual data, and shows alarm status.
IoT Data Analysis Curation
AI analyzes the target IoT data and curates recommended visualizations and analysis directions. It also provides natural language-based explanations of the recommended analysis approaches and data, and responds to user questions.
Automated Recommended Analysis
Automated recommended analysis helps users make more effective decisions by providing customized data analysis and insights. This allows users to quickly grasp the core of the data and improve business performance.
Analysis Flowchart
The above analysis flowchart visualizes the data analysis sequence, making it easy for users to understand and approach the analysis.
AI Analysis Recommendations
AI provides explanations of the analysis recommendations based on the data. Additionally, users can ask questions to receive explanations about the data and analysis.
IoT Data Visualization and Analysis 1
Time Series Data Analysis
• Identify performance trends of equipment over time
• Detect abnormal sudden changes or outliers
• Discover periodic patterns and repetitive behaviors
Histogram Analysis
• Assess the normal operating range of sensor values
• Analyze data skewness and identify outliers
• Determine threshold values for alarm settings
Box Plot Analysis
• Compare performance differences by time of day/day of week
• Analyze the impact of seasonal changes on equipment
• Identify unusual patterns that occur under specific conditions
Heat Map Analysis
• Discover performance change patterns at specific times/days
• Analyze differences in operation between day/night, weekday/weekend
• Provide a basis for adjusting operating hours to optimize performance
IoT Data Visualization and Analysis 2
Daily/Hourly Pattern Analysis
• Identify recurring daily/weekly patterns
• Predict equipment load at specific times
• Optimize production planning and energy usage
Anomaly Detection
• Early detection of equipment anomalies
• Provide opportunities for preventive action before failures
• Optimize maintenance schedules to reduce costs
Frequency Analysis (FFT)
• Analyze equipment vibration characteristics to diagnose wear
• Identify hidden periodic patterns
• Separate noise from meaningful signals
Time-Frequency Analysis
• Track changes in frequency characteristics over time
• Understand relationships between transient events and frequency patterns
• Identify time-frequency characteristics of abnormal operating conditions
IoT Data Visualization Analysis 3
Trend Analysis
• Identify long-term performance trends of equipment
• Predict future performance and prepare for seasonal variations
• Detect early signs of gradual performance degradation
Correlation Analysis
• Identify key factors influencing sensor values
• Derive core variables for optimizing equipment operation
• Quantitative analysis of cause-effect relationships
Comparative Group Analysis
• Verify the effects of equipment improvements/changes
• Compare performance under diverse operating conditions
• Analyze the distribution differences between normal and abnormal states
3D Visualization Analysis
• Intuitively grasp complex multi-variable relationships
• Discover hidden data clusters
• Identify outliers in the 3D space
IoT Data Visualization Analysis 4
Pattern Clustering Analysis
• Automatic grouping of similar daily/weekly patterns
• Identification and classification of equipment operation modes
• Separation of normal and abnormal pattern clusters
Alarm and Threshold Setting
• Recommended optimal thresholds based on data
• Comparison of fixed/dynamic/outlier-based alarm strategies
• Improved alarm system reliability by minimizing false alarms
Operation/Shutdown Cycle Analysis 1
• Detection of operation/shutdown cycles
• Detection of abnormal cycles and values
Operation/Shutdown Cycle Analysis 2
• Extraction and visualization of abnormal cycles and values
Prototyping Demo Video
Core Values
DX-AI IoT Prism provides the following core values to help users maximize the use of IoT data.
Intelligent Interaction
Through GenAI-based natural language question-answering and analytical recommendation features, even non-data experts can easily explore data, perform necessary analysis, and discover hidden insights.
In-depth Analysis Support
Beyond simple visualization, it provides various analysis modules such as time series analysis, distribution verification, correlation analysis, and anomaly detection, supporting the exploration of the hidden meaning in the data and in-depth analysis.
Efficient Decision Making Support
It presents data-based analysis results clearly, enabling users to make faster, more accurate, and information-based decisions.
Data Integration & Accessibility
It integrates dispersed IoT data in one place and increases data accessibility through an intuitive interface, making it easier to start the analysis.
Intuitive Visualization
It provides complex data in various easy-to-understand visualization forms, helping to quickly identify data patterns, trends, and anomalies, and gain insights. The consistent Prism theme enhances the user experience.
Key Features & Future Opportunities
DX-AI IoT Prism goes beyond simple data visualization, providing core capabilities for strategic utilization of IoT data and intelligent analysis, while also encompassing future expandability. This platform accelerates the digital transformation of enterprises by converting the vast amount of IoT data generated in industrial settings into valuable business insights. It integrates information collected from diverse data sources, visualizes it intuitively, and supports data-driven decision-making through advanced analytics powered by AI technology.
4
Strategic Value Creation
Business innovation and decision support
Intelligent Analysis
AI-driven insights
Data Visualization
Intuitive data representation
4
Data Integration
Collection of IoT data from various sources
The functional architecture of DX-AI IoT Prism is designed as a value pyramid, where each stage builds upon the previous one to create higher levels of value:
1. Data Integration
Reliably collects IoT data from various industrial sensors, devices, and systems, and converts them into standardized formats. It supports both real-time data streams and batch processing, and includes data quality management and metadata management capabilities.
2. Data Visualization
Visualizes the collected data in various chart, graph, and dashboard formats, allowing for intuitive understanding of complex information. It provides customizable visualization templates, interactive dashboards, and multi-dimensional analysis views to effectively identify data patterns and trends.
3. Intelligent Analysis
Utilizes machine learning and AI algorithms to uncover hidden patterns in the data and detect anomalies. Through advanced analytics such as predictive analysis, anomaly detection, and process optimization, it derives actionable insights from the data. A natural language-based question-answering system enables users to easily access and analyze the data.
4. Strategic Value Creation
Ultimately, it links the analyzed insights to business decision-making to generate tangible value. It supports the achievement of strategic organizational goals, such as improving operational efficiency, reducing costs, developing new business models, and enhancing customer experience. It fosters a data-driven decision-making culture and enables continuous innovation.
Beyond its current capabilities, DX-AI IoT Prism has various expandable functionalities and development directions, allowing it to evolve as a flexible platform that can adapt to the changing future industrial landscape.
Expanding Opportunities for IoT Data Utilization
Improved Data Integration and Accessibility
We provide a centralized platform to integrate and manage IoT data from various sources, allowing users to easily access and analyze the data. This helps to eliminate data silos and enhance enterprise-wide data utilization.
Increased Operational Efficiency
By real-time monitoring and analysis of equipment status, production processes, and environmental data, we can detect anomalies early and enable proactive maintenance, minimizing downtime and maximizing operational efficiency.
New Business Value Creation
By analyzing the accumulated IoT data, we can uncover new business opportunities for product improvements, service development, and enhanced customer experiences, driving data-driven innovation.
GenAI-based Analysis and Q&A
Natural Language Interface
Users can ask questions about data and request analyses in natural language (Natural Language) without complex queries or coding. For example, the system can respond to questions like "Show me the temperature change trend of a specific facility last week" or "What is the most frequently occurring alarm type?" and provide relevant visualizations.
Automatic Insight Generation
GenAI models can automatically identify and summarize subtle patterns, correlations, and anomalies in large datasets that humans may easily miss, adding depth to data analysis.
Support for Generating Analysis Reports
Based on the analysis results, the system can automatically generate summary reports or explanatory text, reducing the user's document creation burden and facilitating the sharing of analysis results.
Intelligent Analytics Recommendations
Data Collection
Collect data from various IoT sources
Understand Data Characteristics
Analyze data types such as time series, distribution, categorical, etc.
Recommend Analysis Methods
Propose analysis techniques optimized for the data
Provide Visualization
Present results using appropriate visualization types
Data-driven Recommendations: The system identifies the characteristics of the data being analyzed (e.g., time series, distribution, categorical) and recommends the most suitable analysis methods or visualization types (e.g., trend analysis, correlation analysis, box plot) for the user.
Understand User Intent: Based on the user's queries or previous analysis history, the system infers the analysis goals and suggests relevant additional analyses or data items to explore, helping to uncover deeper insights.
Future Opportunities
Advanced Predictive Analytics
By combining GenAI and machine learning models, we can develop more sophisticated predictive models (e.g., remaining life prediction, quality prediction) and improve the credibility of the predictions by explaining the results in natural language.
Autonomous Operations and Control
Based on the analysis results, the system can autonomously determine the optimal operating conditions and adjust the equipment control parameters, evolving into an automated and autonomous operation system.
Personalized Analytics Experience
By providing personalized dashboards and analytical recommendations tailored to the user's role, interests, and analysis level, we can maximize user satisfaction and utilization.
External Data Integration
By combining IoT data with external data sources such as public data, weather information, and market trends, we can perform more comprehensive and multifaceted analyses, and generate new insights.
Future Considerations
The following considerations can be further explored to continuously enhance and expand the capabilities of the DX-AI IoT Prism system.
1
Large-scale IoT Data Processing Architecture
Address the increase in data volume due to the growing number of sensors and reduced data sampling intervals.
2
Real-time Monitoring and Alert System
Improve the ability to detect critical events by applying intelligent anomaly detection algorithms.
3
Embedded Machine Learning Models
Establish an environment for direct model training and management within the system.
4
Advanced Analytics Capabilities
Deepen analytical capabilities by adding domain-specific analysis functions.
Advanced Real-time Monitoring and Intelligent Alert System
Intelligent Anomaly Detection
Going beyond simple threshold-based alarms, the application of intelligent anomaly detection algorithms that can identify complex conditions or time-series pattern changes can reduce false positives and enhance the ability to detect important events.
Real-time Stream Processing
Leveraging stream processing technology, the system should support real-time analysis and alert generation as data is generated, and strengthen the alert notification functionality across various channels (SMS, email, messenger, etc.).
Internalization of Regression/Classification Model Training and Prediction Capabilities
We consider building an MLOps (Machine Learning Operations) environment where regression, classification, and other machine learning models can be trained and managed directly within the system.
Through a standardized interface, users can easily train, evaluate, and deploy models, as well as utilize real-time prediction services. (e.g., integration with scikit-learn, TensorFlow/PyTorch)
Model Training Interface
Provides an intuitive interface for users to select datasets and adjust model parameters to train machine learning models.
Model Evaluation Dashboard
Evaluates the performance of trained models using various metrics and visualizes the results to validate the model's reliability.
Model Deployment Pipeline
Builds an MLOps pipeline to deploy verified models as real-time prediction services and monitor their performance.
Expansion of Various Advanced Analytics Capabilities and Automated Report Generation
Expansion of Various Advanced Analytics Capabilities
In addition to the current visualizations provided, advanced analytics capabilities tailored to specific domain requirements, such as root cause analysis, predictive maintenance models, and process optimization algorithms, can be added to deepen the analysis capabilities.
We are considering the introduction of an extensible plugin architecture that allows users to directly add and execute analysis scripts or models.
Automated Data Analysis and Report Generation
By further enhancing the GenAI functionality, we can automatically generate basic exploratory data analysis (EDA) results and summarize key features when data is loaded.
We will implement a feature to automatically generate and distribute customized analysis reports based on user-defined schedules and conditions, automating repetitive analysis and reporting tasks.
Podcast about IoT Prism