During the project, we've developed a software app to help you estimate the corrosion risk based on water quality parameters. In this blogpost, you can find the instructions on how to use the software.
Introduction
The SOCORRO App is an online tool developed within the framework of the Interreg2Seas SOCORRO project. The application is based on an AI model and allows to calculate a corrosion risk based on measured environmental parameters. The data needs to be in a specific format and can be uploaded online. No uploaded data is stored after usage of the application, nor can it be retrieved. The model calculates the corrosion risks, which can be downloaded as a csv file.
The link to the online tool can be found here.
The parameters used by the model to calculate the corrosion risk are:
Date and Time (for Deep Learning based timeseries models)
Temperature [°C]
DO, Dissolved Oxygen [mg/l]
pH [-]
ORP [mV], Oxidation-Reduction Potential
Conductivity at 25 [µS/cm]
The application is straightforward to be used. How to use the application is explained in this instruction video.
File format
A specific file format needs to be respected when uploading the data. The data should contain 23 columns. The algorithm checks the headers for the presence of the following strings:
Meas_Rate, in column 2
Date, in column 4
Time, in column 5
DO, in column 6
pH, in column 10
ORP, in column 12
Conductivity at 25, in column 16
This file should be a csv file, not an excel file (to avoid formatting issues with different users having different excel settings). Use ';' as delimiter.
Date format: dd/mm/yyyy (This means that Januari 5th 2022 is written as: 05/01/2022 (i.e. not 5/1/2022!)
Time format: hh:mm:ss (This means for example: 05:07:35; not 5:7:35)
Only lines that have a correct input in the columns with headers as mentioned above will be retained for calculation of the corrosion risk.
Type of model to be selected
For calculation of the corrosion risk, 5 different models can be selected, as can be seen in the print screen below.
The table below provides the type of model and material grade of which data was used for training of the model. The Deep Learning models expect a continuous timeseries as input, as they are models that take into account autoregressive effects (I.e. history of the parameters). The regression model will not make use of the date and time of the individual data points.
