Open source experiences in AI and DataViz for manufacturing market
2 likes134 views
1) The document discusses three cases using AI and data visualization in smart manufacturing. The first case involves real-time monitoring and early warning systems using anomaly detection and forecasting. The second case uses connected car and spare parts history data to predict future operations and costs. The third case applies algorithms like clustering and decomposition to disaggregate building energy consumption data from appliances. 2) Grazia Cazzin leads the Knowage Labs and Data & Analytics competency center at Engineering Group, with experience in enterprise applications, data warehousing, BI, and big data analytics for industries, finance, and public sectors. 3) Usability and reducing complexity are emphasized for effective data visualization and AI applications.
Open source experiences in AI and DataViz for manufacturing market
- 1. Open source experiences in #AI and #DataViz for manufacturing market DATA & ANALYTICS CoE
- 2. About me Introduction Grazia Cazzin is the offering manager for Big Data & Analytics competency center in the Research and Innovation Division of Engineering Group S.p.A.. She is the Knowage Labs Director (www.knowage-suite.com), the founder of the SpagoBI project (http://www.spagobi.org) and its historical project leader. Working in the IT field since 1992, she has been involved in enterprise application development, data modelling, data warehousing, dimensional analysis and business intelligence; she has gained valuable expertise working in several market sectors (industry, finance, public administration) and covering several thematic areas (ERP, MRP, MPS, Enterprise Portals, CRM, DWH and BI, Big Data, Analytics). She's currently involved in all the activities concerning Big Data Analysis, as for SmartCity, Smart Agriculture and Industry 4.0 scenarios.
- 4. Data & Analytics compentecy center Introduction
- 5. #case 1: Early warning & RULE estimation AI & Dataviz in smart manifacturing • Real Time • Current data only Monitoring Early Warning Predictive Maintenance • Near real-time • Current data and recent history • Deferred system • Recent data and long history • Real Time • Monitoring • Alarm • Early Warning On plant Data Lake Support • Data Storage • Prediction • Anomaly Detection • Impact analysis • Evaluation Dashboards • Scenario Simulations • Strategic decision • Maintenance Schedule Registry Feedbacks Additional information Time Line object X Features Jobs Alarms State Sensors Time-based data Logical architecture Analytical layer
- 6. #case 1: Early warning & RULE estimation AI & Dataviz in smart manifacturing Data preparation Text mining Feature mapping featuressensors text messages Real-Time monitoring Deployed in plant Early Warning Aggregation Forecasting Anomaly detection Un-supervised learning Predicted states Fault likelihood Deviation index Anomaly detection Un-supervised learningAnalytical flow
- 7. #case 1: Early warning & RULE estimation AI & Dataviz in smart manifacturing
- 8. #case 1: Early warning & RULE estimation AI & Dataviz in smart manifacturing Make it easy … to use !
- 9. #case 2: connected car & spare parts mng AI & Dataviz in smart manifacturing Looking at general attributes of every operations And using the history of everyone Predict next operation, involved spare parts and related costs And the next ones, setting the future history of operations that will be out of the official network HISTORY FUTURE Providing useful information for retention campaigns • Drive-style from connected car • Spare parts semantic map • Fault history (operation and SP) • KB analysis (text analysis) • Prediction • Impact on sales (dashboards)
- 10. #case 2: connected car & after sales AI & Dataviz in smart manifacturing Make it easy … to use !
- 11. #case 3: building energy consumption AI & Dataviz in smart manifacturing light line robot computer HVAC alarm HVAC light lamp alwaysON Appliance Signature identification De-noising Filtering State detection Clustering of devices Probability distribution function estimation Disaggregation Algorithms From signatures to appliances combination
- 12. #case 3: building energy consumption DATA & ANALYTICS CoE Feature Extraction Consumption Distribution Updating device statics Output Device Identification Event Detection Data Preprocessing Data Transformation Data Integration Data Reduction Data Cleaning Wavelet filter - Data Normalization - Missing data reconstruction Data Acquisition Fourier transform Noise reduction Apache Spark Cassandra Hive MySQL Input Peaks and valley detection spike removal Steady state model Signal decomposition Temporal behavior detection Laplace transform Wavelet transform Clustering Event Pairing ON OFF Self Organizing Map DBSCAN / HDBSCAN KMEANS KNN Signal Similarity Distance Meanshift Fuzzy C-means K-medoids Dynamic Programming Kernel Density Estimation Fuzzy rule learner Random Forest Gaussian Mixture Model Generative Adversarial Network Masked autoencoder for distribution estimation Smart Device ranking Fuzzy Reasoning Adaptive probabilistic graphical model User experience
- 13. #case 3: building energy consumption DATA & ANALYTICS CoE Make it easy … to use !
- 14. Last mile: DataViz as AI enabler Conclusion #Usability#Complexity
- 15. Conclusion @gcazzin Grazia Cazzin www.knowage-suite.com www.eng.it THANK YOU ! Meet us at Booth A3