Alarm Monitoring

Alarm classification

Input Filtering

During the “SCADA years”, the documentation was not in line with the prices charged for the product. We also felt that the worts were starting to show on the aging Microsoft platforms that were periodically failing at some of the largest installations in Australia. If you have some ideas for a SCADA product, contact us for a discussion. We had been working on several systems with very high data rates and wanted to visualise massive datasets. Unfortunately, these things require funding, so if there is a potential product even if partially open-sourced, we would be willing to commit resources to it. Take a look at what we think documentation for various thresholds should look like, but we welcome your suggestions.


Figure 1: Alarm Classification. Alarm monitoring can be configured to run on intelligent input cards, or could run on a host that periodically receives the raw input value.

Alarm High Thresholds

There are a lot of LaTeX files with high resolution graphics on debouncing digital inputs, these alarm diagrams (higher resolution on the PDF versions), and what we think SCADA screens should look like. As a hint, we think input for the background (or static components) should be done in Photoshop and Illustrator. Web technologies on Linux/ Mac OS X or Unix will handle “hot-spots” but the idea of a “thick client” on a tablet is even more promising. If you get a chance, test some OpenGL code on an iPad to see how impressive even the original 1 GHz A4 processor handles 3D plots. We may post the LaTeX output when we have split the chapter versions into article sized chunks. The diagrams were created in a student version of Illustrator that my daughter used, and in 2014 the Cloud versions of Adobe became much more affordable.


Figure 2: The analog alarm values can be configured to the intelligent input cards, or could run on a host that periodically receives the raw input value.

Alarm Low Thresholds


The analog low threshold diagram shows how the raw values are preprocessed to minimise ‘nuisance trips’.

Figure 3: Similar to the above plot, but showing the low— and low-low analog values.

Analog Alarm States


Figure 4: Possible transitions assuming sufficient sampling rates. Skipping a state could indicate a catastrophic failure.

Alarm State Diagram


Figure 5: State transition diagram for a monitoring task in a real-time kernel on an embedded target with the alarm processing done at the sensor/ controller interface.

When the task is initialised, it determines which state to enter based on the analog value and assumes the state was entered from normal. The simple diagram does not show mode transitions, which occur when starting up a plant — similar to avionics when the plane transitions from “take off”, to “cruize” and then to “landing”.