100% leak recall and above 80% leak accuracy on a produced water pipeline.
100% leak recall and above 80% leak accuracy on a produced water pipeline.
100% leak recall and above 80% leak accuracy on a produced water pipeline.
How a midstream produced-water operator validated DPx performance across changing pump configurations. Achieving full leak recall with stable, low-noise alarming.
How a midstream produced-water operator validated DPx performance across changing pump configurations. Achieving full leak recall with stable, low-noise alarming.
Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
Optimal effective coverage determined at 1–2 miles (2–3 km) per sensor
Optimal effective coverage determined at 1–2 miles (2–3 km) per sensor
Optimal effective coverage determined at 1–2 miles (2–3 km) per sensor
Adaptive filters improved signal isolation under higher-noise, two-pump conditions
Adaptive filters improved signal isolation under higher-noise, two-pump conditions
Adaptive filters improved signal isolation under higher-noise, two-pump conditions
100% recall using the aggregation model, with no missed leaks
100% recall using the aggregation model, with no missed leaks
100% recall using the aggregation model, with no missed leaks
Case study profile.
Case study profile.
Case study profile.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
A Canadian midstream operator conducted a live field trial to validate DPx performance under real pipeline conditions. The goal was to confirm reliable detection on a short, difficult-to-instrument asset where traditional metering and conventional leak detection were proving challenging.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
Asset type — Produced-water lateral.
Length/size — 1 mile (1.6 km), 2-inch line.
Operating scenarios — one-pump operation (26 m3/hr.) and two-pump operation (35 m3/hr).
Leak size tested — 3–5% of flow.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
Asset type — Produced-water lateral.
Length/size — 1 mile (1.6 km), 2-inch line.
Operating scenarios — one-pump operation (26 m3/hr.) and two-pump operation (35 m3/hr).
Leak size tested — 3–5% of flow.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
Asset type — Produced-water lateral.
Length/size — 1 mile (1.6 km), 2-inch line.
Operating scenarios — one-pump operation (26 m3/hr.) and two-pump operation (35 m3/hr).
Leak size tested — 3–5% of flow.
The pipeline’s configuration introduces operational complexity and elevated consequence in the event of a leak, placing high demands on detection sensitivity and reliability.
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
- Maintain detection performance under one-pump and two-pump conditions
- Avoid missed events while minimizing false alarms
- Adapt to changing pump noise without repeated model retraining
- Establish practical sensor coverage range per installation
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
- Maintain detection performance under one-pump and two-pump conditions
- Avoid missed events while minimizing false alarms
- Adapt to changing pump noise without repeated model retraining
- Establish practical sensor coverage range per installation
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
- Maintain detection performance under one-pump and two-pump conditions
- Avoid missed events while minimizing false alarms
- Adapt to changing pump noise without repeated model retraining
- Establish practical sensor coverage range per installation
The objective was to quantify DPx leak detection accuracy, recall, and adaptability under varying pump configurations and flow conditions during live operations. The operator needed to validate whether an acoustic edge system could deliver dependable leak detection in live service, specifically:
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
* Instantaneous detection model — evaluates leak probability every 0.1 seconds for rapid response and visibility
* Aggregation detection model — applies a rolling window with a 50% threshold to stabilize alarming and minimize false positives
The ML model architecture and weights remained constant throughout the trial. Only frequency cutoff bands were tuned on-site to adapt to pump noise.
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
* Instantaneous detection model — evaluates leak probability every 0.1 seconds for rapid response and visibility
* Aggregation detection model — applies a rolling window with a 50% threshold to stabilize alarming and minimize false positives
The ML model architecture and weights remained constant throughout the trial. Only frequency cutoff bands were tuned on-site to adapt to pump noise.
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
* Instantaneous detection model — evaluates leak probability every 0.1 seconds for rapid response and visibility
* Aggregation detection model — applies a rolling window with a 50% threshold to stabilize alarming and minimize false positives
The ML model architecture and weights remained constant throughout the trial. Only frequency cutoff bands were tuned on-site to adapt to pump noise.
Pipewise deployed DPx, a compact edge-based leak detection system combining high-frequency dynamic-pressure sensing, on-device ML-based detection, and adaptive filtering tuned on-site to isolate leak signatures under different pump noise conditions. Two ML-based detection approaches were evaluated:
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
* ≥80% detection accuracy in both one-pump and two-pump scenarios using the instantaneous model
* 100% recall using the aggregation model, with no missed leaks
* Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
* Adaptive filters improved signal isolation under higher-noise, two-pump conditions
* Optimal effective coverage determined at ~1–2 miles (2–3 km) per sensor
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
* ≥80% detection accuracy in both one-pump and two-pump scenarios using the instantaneous model
* 100% recall using the aggregation model, with no missed leaks
* Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
* Adaptive filters improved signal isolation under higher-noise, two-pump conditions
* Optimal effective coverage determined at ~1–2 miles (2–3 km) per sensor
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
* ≥80% detection accuracy in both one-pump and two-pump scenarios using the instantaneous model
* 100% recall using the aggregation model, with no missed leaks
* Improved precision at the mid-line sensor location as low-frequency pump noise attenuated
* Adaptive filters improved signal isolation under higher-noise, two-pump conditions
* Optimal effective coverage determined at ~1–2 miles (2–3 km) per sensor
DPx delivered consistent, measurable performance across operating conditions — demonstrating high sensitivity, full recall, and stable alarming under real-world variability.
Why it matters.
Why it matters.
Why it matters.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
By combining high-frequency sensing, edge-based ML, and adaptive filtering, DPx closes a critical compliance and risk-management gap for operators managing challenging assets.
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
* Short laterals and stub lines
* Produced-water gathering segments
* High-consequence areas (HCAs)
* Pipeline sections with limited SCADA or metering infrastructure
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
* Short laterals and stub lines
* Produced-water gathering segments
* High-consequence areas (HCAs)
* Pipeline sections with limited SCADA or metering infrastructure
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
* Short laterals and stub lines
* Produced-water gathering segments
* High-consequence areas (HCAs)
* Pipeline sections with limited SCADA or metering infrastructure
This field validation confirms DPx is well-suited for assets that are difficult to monitor with traditional instrumentation or conventional leak detection systems, including:
Need reliable leak detection on short laterals, stub lines, or high-noise assets?
