Differences

This shows you the differences between two versions of the page.

--- ip:ws2025:do_sensor:start [2026/02/10 09:15] – [Alexander Rausch (33213)] 31629_students.hsrw
+++ ip:ws2025:do_sensor:start [2026/02/25 07:10] (current) – 32603_students.hsrw
@@ Line 14: / Line 14: @@
 [[ https://gitlab.hsrw.eu/31507/ip_25 | Project Repository ]]
+[[ https://gitlab.hsrw.eu/32603/ip25-esp32-sketch | ESP32 Sketch ]]
 {{ :ip:ws2025:do_sensor:ip-25_final_project_report.pdf |PDF-Version of the report}}
@@ Line 575: / Line 577: @@
-=== 2.5.9 Limitations, remaining Issues and possible Improvements (Shamsunnahar, Sultana) ===
+=== 2.5.9 Reference-Based Calibration and Bias Analysis (Müsch) ===
+**2.5.9.1 Motivation and Relation to the System Calibration Concept**
+The prototype evaluation confirms that the dissolved-oxygen (DO) sensing chain is functional and repeatable under laboratory conditions. However, quantifying accuracy requires direct comparison against a certified reference instrument across relevant operating regimes (e.g., different hydrodynamic states and mesh configurations). In addition, the manufacturer recommends a two-point field calibration (100% saturation and a zero-oxygen solution) as a baseline for routine maintenance. Within the project scope, a full two-point procedure and broad controlled sweeps over the full DO range were not feasible.
+For the intended long-term deployment with copper mesh, a second question becomes operationally critical: Does the mesh (and the local hydrodynamics around it) introduce a systematic bias, and can this bias be compensated by a simple calibration function that is feasible for embedded implementation?
+Therefore, this section documents a reference-based calibration experiment intended to:
+  * quantify mesh/flow-dependent bias and
+  * derive practical calibration parameters for field deployment.
+**2.5.9.2 Experimental Setup and Tested Water Samples**
+A reference probe (WTW FDO 925) and the project probe were mounted in the same reservoir and measured in parallel over time. The project probe ran custom firmware to enable precise logging of sensor readings and event markers (mesh changes, flow regime changes, and cooling steps to maintain approximately 20°C throughout the experiments).
+<imgcaption image_calib_mounting|>
+{{ :ip:ws2025:do_sensor:img_5307.jpg?nolink&700 | Mounting of reference and project probe in Sample 2 (Tap-Water).}}
+</imgcaption>
+**Logging-Intervals**
+  * Project sensor (SEN0680): logged every **5 seconds**
+  * Reference instrument (WTW FDO 925): read approx. **every minute** and entered near-synchronously into the same CSV record via serial command:
+<code>
+r <DO_READING> <TEMP>
+</code>
+To induce DO dynamics, dissolved oxygen was increased periodically (every 5 minutes) using controlled stirring (magnetic stirrer), followed by quasi-steady segments (15 minutes) in which DO gradually decayed.
+**Configurations**
+Three sensor configurations:
+  * **none**: no mesh
+  * **mesh1**: front-open mesh variant
+  * **mesh2**: full-coverage mesh variant
+Three flow regimes:
+  * **off**: no intentional flow (stagnant)
+  * **low**: gentle stirring (approximate laminar mixing)
+  * **med**: increased stirring intensity
+**Water samples**
+^ Sample ^ Description ^ Volume ^ Typical DO ^
+| Sample 1 | BSB water / “Biological Oxygen Demand” | 1600 mL | ~9 mg/L |
+| Sample 2 | Tap water | 3500 mL | ~6–8 mg/L |
+| Sample 3 | Mussel water, aerated | 3000 mL | ~8–9 mg/L |
+//Table 6.1: Overview of water samples used for reference-based calibration and bias analysis.//
+**Operational observation (Sample 3, medium flow):**
+At medium stirring intensity, the vortex generated by the magnetic stir bar entrained air bubbles and transported them preferentially to the reference probe. This can cause short transient phases in which the reference responds earlier than the project sensor despite similar positioning. Consequently, quantitative calibration fitting was prioritised on quasi-steady segments rather than immediate post-aeration mixing transients.
+**2.5.9.3 Calibration Model and Data Processing**
+To keep calibration explainable and suitable for embedded deployment, a linear model was selected:
+<code>
+DO_ref = m · DO_proj + c
+</code>
+where:
+  * **DO_ref** is the reference DO in mg/L (WTW FDO 925)
+  * **DO_proj** is the project sensor reading (scaled to mg/L)
+  * **m** and **c** are slope and offset estimated via least-squares fitting
+For operational interpretation, bias was computed per configuration and flow regime as:
+<code>
+ΔDO = DO_proj − DO_ref
+</code>
+Bias was summarised as **mean ± standard deviation (SD)** to separate systematic over-/underestimation from variability.
+**Segment selection principle (important for robust fitting):**
+  * Prefer quasi-steady phases (decay / stable plateaus).
+  * Down-weight or exclude short-lived transients immediately after aeration, especially under **med** flow (bubble/vortex artifacts).
+**2.5.9.4 Results: Influence of Mesh and Flow (Bias)**
+Across all three samples, the bias analysis indicates that both mesh configuration and flow regime influence measurement behaviour.
+  * In stagnant conditions (**off**), local gradients and boundary-layer effects are expected to be stronger, which can increase both bias and variance.
+  * In contrast, gentle flow (**low**) improves mixing near the membrane and reduces local gradients, typically improving agreement with the reference.
+**Key patterns observed:**
+  * **mesh2 + low flow** produced the most balanced behaviour across samples, with bias values closest to zero and comparatively low variability.
+  * **no-mesh** more frequently exhibited **positive bias** (tendency to overestimate DO), and in some cases higher variability.
+  * **medium flow** can yield good agreement during steady phases, but is more susceptible to transient artifacts during aeration and bubble transport (notably in Sample 3).
+<imgcaption image_bias_sample1|>
+{{ :ip:ws2025:do_sensor:report_mesh_influence_bsb_water_probe1.png?nolink&700 | Bias summary (mean ± SD) across mesh and flow regimes for Sample 1 (BSB-Water). Source: Own analysis.}}
+</imgcaption>
+<imgcaption image_bias_sample2|>
+{{ :ip:ws2025:do_sensor:report_mesh_influence_tap_water_probe2.png?nolink&700 | Bias summary (mean ± SD) across mesh and flow regimes for Sample 2 (Tap-Water). Source: Own analysis.}}
+</imgcaption>
+<imgcaption image_bias_sample3|>
+{{ :ip:ws2025:do_sensor:report_mesh_influence_mussel_water_probe3.png?nolink&700 | Bias summary (mean ± SD) across mesh and flow regimes for Sample 3 (Mussel-Water). Source: Own analysis.}}
+</imgcaption>
+<imgcaption image_calib_timeseries_sample1|>
+{{ :ip:ws2025:do_sensor:report_plot_bsb_water_probe1.png?nolink&700 | Time series comparison (reference points vs. project sensor) for Sample 1 (BSB-Water). Source: Own analysis.}}
+</imgcaption>
+<imgcaption image_calib_timeseries_sample2|>
+{{ :ip:ws2025:do_sensor:report_plot_tap_water_probe2.png?nolink&700 | Time series comparison (reference points vs. project sensor) for Sample 2 (Tap-Water). Source: Own analysis.}}
+</imgcaption>
+<imgcaption image_calib_timeseries_sample3|>
+{{ :ip:ws2025:do_sensor:report_plot_mussel_water_probe3.png?nolink&700 | Time series comparison (reference points vs. project sensor) for Sample 3 (Mussel-Water). Source: Own analysis.}}
+</imgcaption>
+The full results from the lab experiments including the source codes for logging (during experiments) and analysis are provided for future reference:
+  * {{ :ip:ws2025:do_sensor:experiment_analysis.zip | Ressources used for analysis (Archive)}}
+**2.5.9.5 Recommended Calibration for Field Deployment (mesh2)**
+The intended field scenario prioritises anti-fouling robustness; therefore **mesh2** is the most likely deployment configuration. Based on the reference-based fits and robustness across samples, the following strategy is recommended:
+  * If flow classification is feasible, a **two-regime approach** (**low** vs. **off**) is preferred because it preserves accuracy under stagnant conditions while keeping the model simple.
+  * Otherwise, a **single-curve approach** is acceptable when operational simplicity is prioritised.
+^ Regime ^ Calibration function ^
+| mesh2 / low | DO_ref ≈ 1.249 · DO_proj − 1.763 |
+| mesh2 / off | DO_ref ≈ 1.238 · DO_proj − 1.673 |
+| mesh2 / all | DO_ref ≈ 1.231 · DO_proj − 1.597 |
+//Table 6.2: Example calibration functions derived from the dataset for mesh2.//
+**Note:** These parameters are derived from laboratory measurements (three samples at controlled temperature and defined stirring regimes). Field conditions may differ (biofouling, temperature swings, real turbulence). A dedicated field validation phase is recommended.
+**2.5.9.6 Interpretation and Operational Notes**
+**Mesh-related effects:**
+The copper mesh modifies the local hydrodynamics around the sensing membrane. The results indicate that **mesh2** can be calibrated to high short-term accuracy and that **gentle flow** reduces boundary-layer effects. Importantly, the mesh does not introduce unacceptable short-term noise in the presented experiments, provided that the hydrodynamic regime during calibration and operation is broadly comparable.
+**Flow-related effects:**
+Gentle flow improves agreement with the reference by reducing local gradients and improving exchange at the membrane interface. Medium flow can be acceptable during steady phases, but may amplify transient artifacts during aeration and bubble transport (vortex entrainment), which is particularly relevant for laboratory mixing setups.
+**Conservative monitoring perspective (alerts vs. calibration):**
+For environmental monitoring, missing low-oxygen events is typically more harmful than triggering occasional false alarms. Any safety margin should be implemented in the **alert logic** (e.g., slightly lower alarm thresholds or persistence-based alarms) rather than by intentionally biasing the calibration function, which would reduce comparability and interpretability of long-term time series.
+**Maintenance:**
+Depending on pollution and suspended solids, the copper mesh can clog with dirt and organic material (observed during Sample 3). Reduced permeability and accumulation around the mesh can impair exchange near the sensing surface, leading to drift and/or reduced responsiveness. Periodic inspection and cleaning intervals should therefore be defined based on deployment conditions.
+**2.5.9.7 Limitations**
+  * The experiment uses three representative samples but does not cover the full 0–20 mg/L DO range; broader controlled sweeps would improve generality.
+  * Temperature was held approximately constant (~20°C). Temperature dependency and compensation should be validated for seasonal field deployment.
+  * Medium-flow phases can introduce short-lived reference-leading artifacts due to bubble/vortex transport; calibration should therefore prioritise quasi-steady segments.
+  * Long-term drift and fouling behaviour under real environmental conditions remain to be validated in extended deployments.
+=== 2.5.10 Limitations, remaining Issues and possible Improvements (Shamsunnahar, Sultana) ===
 The performance of copper-based anti-fouling systems depends on environmental factors such as temperature, pH, salinity, and flow velocity, which influence both copper ion release rates and convective transport. These parameters were not systematically investigated in this study.
@@ Line 583: / Line 736: @@
 Further optimisation of mesh geometry and porosity may improve convective mixing while maintaining antifouling effectiveness. A hybrid approach, combining passive copper protection with occasional low-power mechanical cleaning, could provide increased robustness against both microfouling and macrofouling. In addition, environmentally friendly surface coatings could be explored as complementary measures to further extend maintenance-free deployment duration.
 ----
@@ Line 866: / Line 1017: @@
 Finally, I configured an MQTT node provided by the university to receive real sensor data and visualize it in Grafana. To ensure knowledge transfer and facilitate future work, I wrote a comprehensive README file in the project repository, documenting setup procedures and handover information for a subsequent team.
+=== Mohamed Karim (32603) ===
+{{ :ip:ws2025:do_sensor:timesheet_mohamed_karim.pdf |Timesheet Mohamed Karim (32603)}}
+I was responsible for the complete hardware development of the IP-25 monitoring node,
+including system architecture, sensor integration, power design, and laboratory validation. I
+designed and implemented the full sensing chain around the ESP32 microcontroller,
+integrating the optical RS485 dissolved oxygen probe (DFRobot SEN0680) and the analog
+pressure sensor, including the necessary voltage scaling and electrical protection. I
+developed the high-side switching concept for sensor power control, implemented the
+prototype using a relay-based solution due to component availability, and documented the
+intended MOSFET optimization for the final revision.
+In addition, I designed the entire power management architecture, including charger
+selection, boost converter topology, duty-cycled operation strategy, and a worst-case
+analytical energy budget aligned with the 7-day maintenance constraint. The resulting model
+demonstrated an estimated runtime of approximately 4.9 weeks on a 2,000 mAh LiPo
+battery. I assembled and soldered the prototype hardware, performed end-to-end sensor
+testing under laboratory conditions, evaluated dissolved oxygen readings against saturation
+tables, calibrated the pressure sensor zero offset, and documented all calculations, figures,
+and validation results in the final report.
+According to my documented timesheet, this contribution amounts to approximately 184
+hours of hardware design, implementation, testing, debugging, and technical documentation.
 === Yannick Müsch (31490) ===
@@ Line 874: / Line 1051: @@
 I produced two 3D-print iterations, stitched the mesh carefully to the cage and validated the mechanical fit and usability of the caps. Together with Mohamed, I planned and executed practical tests of both variants and documented the results (including logs). Based on these tests, we could not observe short-term negative effects or interference from introducing the copper mesh. All findings were then compressed into the respective sections of the report.
+Later on, I led a more comprehensive series of experiments at LINEG to further quantify the effect of different variables such as mesh-configuration and water flow. The experiments were performed in a controlled laboratory setting at approx. 21°C and equal mounting of the reference probe (WTW FDO 925) and project probe (SEN0680). With three different water samples (BSB/"Biological Oxygen Demand", Tap, Aereated Mussel Water) and controlled stirring (using a magnetic stirrer), a DO range spanning from approx. 6 mg/L to 9 mg/L was covered. The in-depth analysis of these results was then used to derive a linear calibration model for the project probe to reduce the error (as measured by RMSE) and correct the measurement bias (i. e. slightly lower readings with full-cover mesh) to resemble the reference probe more closely. The presented results were then documented and discussed.
 === Stefan Rak (31604) ===
@@ Line 898: / Line 1077: @@
 === Johannes Schelb (31861) ===
+{{ :ip:ws2025:do_sensor:timesheet_johannes_schelb_31861.pdf |Timesheet Johannes Schelb (31861)}}
 At the beginning of the project, I actively participated in the initial brainstorming phase, during which fundamental ideas regarding the design of the prototype were developed. In a second brainstorming phase, I worked together with my project group to develop a concrete prototype concept and contributed to the preparation and presentation of the corresponding concept proposal.
@@ Line 917: / Line 1098: @@
 === Lars Theodor Schöne (31629) ===
+{{ :ip:ws2025:do_sensor:lars_schoene_timesheet.pdf |Timesheet Lars Theodor Schöne (31629)}}
 At the beginning of the semester, when the entire course was still working as a single group, I actively contributed to the development of an initial prototype plan. As I have limited technical expertise in hardware and system assembly, I focused primarily on conceptual aspects and the integration of creative ideas for the prototype. This included, for example, the idea of using a pipe-based housing as well as the exploration of different cleaning and maintenance concepts for the measurement station.
@@ Line 940: / Line 1123: @@
 === Fabian Sloma (31647) ===
+{{ :ip:ws2025:do_sensor:timesheet_31647.pdf |Timesheet Fabian Sloma (31647)}}
 During the initial concept phase, I actively contributed to the brainstorming process, focusing on a design concept for a floating, enclosed sensor modeled after a buoy. This early draft included provisions for a protective mesh to shield the sensor head and a preliminary arrangement of specific internal components.
@@ Line 951: / Line 1136: @@
 === Subrina Sultana (33977) ===
+{{ :ip:ws2025:do_sensor:timesheet_subrina_sultana_33977_.pdf | Timesheet Subrina Sultana (33977)}}
 During the project, I mainly contributed to the Cleaning / Anti-Fouling work package, with a strong focus on improving the long-term reliability of dissolved oxygen (DO) sensors used in water monitoring. My primary task was the development and evaluation of a copper-mesh–based anti-fouling solution, which was designed as a passive method to reduce biofouling caused by algae, biofilm, and debris. I was involved in developing the initial design concept together with another team member and was responsible for creating the base design. I contributed to defining the concept’s purpose and analyzing how the copper mesh could act as both a physical barrier and a chemical deterrent without negatively affecting sensor measurements. Special attention was given to the positioning of the mesh to ensure sensor protection while maintaining accurate readings.