ThingWorx Analytics Training: Module 2 Part 1
- December 23, 2022
- 0 replies
- 1103 views
This video is Module 2: Use Case Discussion of the ThingWorx Analytics Training videos. It covers what a use case is, and what a successful use case requires. It details a few examples that have been explored using ThingWorx Analytics.
This video does not contain any hands-on exercises, and you don't need access to a running ThingWorx Analytics environment.
This video has chapters — open the chapter menu in the video player to jump to any section.
Chapter Summaries
0:00 – Introduction: ThingWorx Analytics Use Cases with Matt Hardman (PTC)
Matt Hardman, who leads PTC's AI and analytics in-market team, introduces the session on real-world ThingWorx Analytics customer deployments and the process for building sustainable, production-ready predictive analytics use cases.
0:36 – Framing the Use Case: "Given / Predict / Optimize" Business Problem Method
Explains the "Given [data], I want to understand/predict/optimize [outcome] so that [action]" pattern, using examples like predicting electrical failure for preventative maintenance or scrap prediction, stressing that you solve a business problem rather than adopting AI for its own sake.
2:02 – Data Requirements: Feature Correlation, Time Alignment, Quality & Variation
Covers the guiding principles for data readiness: correlating input features (sensor and MES data) with outcomes over the same time window, handling time zones and daylight savings, "garbage in, garbage out" data quality, sufficient volume, adverse-event variation, and roughly 25 examples per discrete value.
4:41 – Start with End-User Value: Field Service & Manufacturing Scenarios
Argues for designing around the end user consuming the application rather than the technology, illustrated with connected-product field-technician root-cause and truck-roll consolidation, plus manufacturing use cases spanning unplanned downtime, availability, throughput, and quality challenges.
6:18 – Flowserve Case Study: ThingWorx Analytics Platform Overview
Introduces Flowserve, a maker of industrial flow-management pumps for oil, gas, and critical industries using ThingWorx as their connected-product cornerstone, then frames the platform's role in delivering automated, real-time actionable intelligence (referencing Gartner's 26 billion connected devices estimate).
8:01 – ThingWatcher & ThingPredictor: Edge Sensor Analytics Pump Demo
Walks through a Flowserve pump rig retrofitted with six National Instruments sensors and Hewlett Packard Enterprise high-performance computing, showing ThingWatcher learning normal operating conditions automatically and ThingPredictor forecasting failure timelines at 30,000 readings per second per sensor.
10:12 – Live Demo: Simulating Flow & Alignment Failures with Real-Time Alerts
Demonstrates injecting faults by closing a valve to reduce flow and shifting pump alignment, showing ThingWatcher flagging anomalies, dropping efficiency, impeller risk, and triggering real-time notifications and automated phone calls to technicians before failure or downtime occurs.
13:11 – Automotive Reflow Oven: Predicting Critical Alarms (3x Reduction)
Details a tier-one automotive manufacturer predicting airflow and oxygen alarms on assembly-line reflow ovens, using three years of data from 10+ ovens with 50 sensors each aggregated into ~500 features, delivered via a ThingWorx mashup that achieved a threefold alarm reduction (from ~10 to ~3 per day).
15:29 – Contact Lens Manufacturer: Alarm-Pattern Analytics for Bottlenecks
Describes work with a contact lens manufacturer where alarm data and machine-to-machine alarm patterns proved more predictive than process telemetry, addressing bottlenecks that drove up to 30% below-efficiency days and generating hypotheses for human domain experts to investigate.
16:56 – Steel Production: Predictive Maintenance for Welder Downtime
Presents an "up the middle" predictive maintenance case for a steel producer losing 2,800+ minutes yearly to welder failure, using 3.5 years of data across ~240,000 attempts at a ~3% failure rate, aggregated into two-hour intervals to alert maintenance teams to high-risk windows in live production.
18:09 – Electric Wire Manufacturer: Fold Prediction & Lift-and-Shift Architecture
Covers an electric wire manufacturer where a model predicts wire folds 40 feet of spool in advance, with the real value being a scalable architecture that lifts and shifts the solution across four identical production lines and sibling plants.
19:15 – Printed Circuit Boards: Early Bridging-Defect Detection (~90% Accuracy)
Explains a printed circuit board maker supplying critical, high-cost assets, where an early-production model detects bridging defects with about 90% accuracy, enabling preemptive quality control that removes defective parts before adding further economic value and reduces cost of quality and warranty exposure.
21:01 – Printer Service via Exeta: Predicting Print-Head Overheating
Describes a printer service company using PTC's Exeta connected-product application to predict print-head overheating a day in advance, drawing on ~9 months of data and 700+ overheating alarms parsed at the component level to enable predictive maintenance and root-cause analysis for engineering.
22:27 – Conclusion: Three Patterns — Value First, Data Second, Technology Third
Wraps up with the recurring pattern across every case: lead with end-user value, then secure enough clean data, and finally apply the right technology (ThingWorx Analytics), warning that reversing the order tends to produce models without end-user acceptance.