POC: From End Goal
to Stepping Stone
Has your POC become an end in itself? toorPIA seamlessly connects analysis to production operation, enabling rapid POC completion and swift decision-making for production deployment.
The Problem with Traditional POC
There is a critical gap between "we analyzed it" and "we can operate it."
With only two data points, the normal "variation range" is unknown, making it impossible to determine if the difference is significant.
With the normal data distribution understood, statistically significant thresholds can be set for objective anomaly detection.
Collect one normal & one abnormal sample
"Let's just use what we have on hand"
Visualize differences with analysis tool
"We can see the difference visually!"
Conclude "anomaly detection works"
"We should be ready to deploy...right?"
Reports can be delivered, but production deployment is not achievable
Design data acquisition process
Standardize when, where, and how to measure
Continuous normal data accumulation
Collect regularly during daily operations
Statistically grounded threshold setting
Derive detection criteria from normal distribution
Operational verification & continuous improvement
Evaluate false positive & false negative rates
A production-ready anomaly detection system can be built
| Comparison | One-Shot Analysis | Operation Approach |
|---|---|---|
| Normal data points | 1 point | Tens to hundreds |
| Measurement conditions | Varies each time | Standardized & documented |
| Threshold basis | "Looks different" | Statistical significance |
| Reproducibility | Unverifiable | Verifiable |
| Stakeholder explanation | Weak evidence | Scientifically explainable |
toorPIA: Both the Output and the Compass
toorPIA's core value lies in capturing precursory signals in the intermediate region between normal and abnormal states. It also serves as a feedback mechanism that rapidly evaluates whether your preprocessing pipeline is configured correctly through visualization.
Data Acquisition
Sound & vibration sensors
Preprocessing
Filtering / STFT / Normalization
toorPIA
High-dim structure visualization
Anomaly Detection
Deviation from normal distribution
Filter Configuration
High-pass and low-pass filter cutoff frequencies. Set according to the frequency bands of the anomalies you want to detect.
More info / more noise Narrowband
Less noise / info loss risk
STFT Window Size
The window width of Short-Time Fourier Transform. Time resolution and frequency resolution are in a trade-off relationship.
Time res. up / Freq. down Large window
Time res. down / Freq. up
Signal Normalization
Sensor signals fluctuate due to gain adjustments and ambient temperature. Normalizing signal intensity and focusing on spectral shape enables robust detection against disturbances.
Robust / intensity lost Raw
Intensity kept / false pos.
Without a compass like toorPIA, optimizing these multiple parameters is extremely difficult.
Open Source Tools Powering Your POC
The environment built during POC becomes your production platform. We provide a tool stack that enables deployment without additional development.
IF-HUB
IndustryFlow Hub — Data Middleware
Middleware connecting industrial data warehouses (such as PI System) to analytics platforms. Access data through unified REST APIs, with virtual tags (gtags) enabling moving averages, Z-score calculations, and other preprocessing directly via the API.
toorpia
Python Client Library
A Python client that creates basemaps from CSV and audio data, projects new data, and performs anomaly detection. MapInspector provides interactive visualization for intuitive cluster attribute comparison and anomaly score review.
Standardize data acquisition and preprocessing with IF-HUB. Analyze and detect with the toorpia client.
The pipeline built during POC becomes your production precursor monitoring platform.
Roadmap to Production
Not a traditional POC, but a streamlined 4-phase approach designed for production deployment.
Data Acquisition Design
Standardize when, where, and how to measure. Document measurement conditions and establish a reproducible data acquisition process.
Baseline Construction
Continuously accumulate normal data during daily operations and understand the normal variation range. Build a statistically significant basemap.
Detection Logic Design
Optimize preprocessing parameters through toorPIA's feedback loop. Derive thresholds statistically from the normal distribution.
Production & Continuous Verification
Evaluate false positive and false negative rates while continuously updating baselines. The system matures through operation.
POC completion is not the goal — it is the beginning of production operation.
Get Started
Not just analysis outsourcing, but operational design partnership.
We design the optimization of the entire workflow together — from data acquisition to preprocessing, visualization, and detection.
