Skip to main content.

This page’s menu:

FMD: Functional Monitoring and Diagnosis Software

We have developed the FMD Software system to monitor and diagnose parameters in a real-time process.

Diagnosis is the act of detecting and isolating an inconsistency between a device and its intended function.

This can be done to a limited extent by using predefined symptoms (sensor alarm limits), or a pattern derived from empirical data (pattern recognition). Pattern recognition is severely limited by:

  1. The training set cannot include all configurations. The combinatorics with even 20 switches and valves is intractable over a million configurations and most all of them are not in the training set since they are not legal or desirable though some may someday require detection and diagnosis.
  2. The training set cannot provide adequate information to derive the correct Physics. Pattern recognition software is never going to derive Bernoulli's equation from historical data.
  3. The techniques have fundamental limitations. Regression analysis is notorious for deriving polynomial equations of the wrong degree, e.g. fitting a 5th degree polynomial when the answer is a square root, due to noisy data. It really only becomes reliable when it is well supervised commercial "pattern recognition" software packages are not only unsupervised, but do not even let the users inspect the derived equations for a sanity check.

The definitive rigorous and reliable basis for diagnosis is the actual operating model.

  1. The configuration data is correct because it is based on the schematic.
  2. The equations are correct because they are the known Physics models.
  3. Some model tuning can safely be performed, in accordance with normal engineering practices, to make slight adjustments to model coefficients in order to better fit actual data but the equation itself is already known and does not need to be discovered by unsupervised software.

By exploiting the operating model, we have the capability to determine errors earlier. Simplifying somewhat, two sensors could have a value in a valid operating range between 0 and 100. In a parametrically based system, there is no inconsistency. But in a functionally based system like FMD, the underlying relationship model says that the two sensors should be equal to certain model values, subject to their tolerances. If this relationship does not hold, we can send an alert to the operator.

In existing monitoring systems, these relationships must be explicitly coded. In the FMD system, these relationships are automatically dynamically derived as needed from the operating model. As changes are made to the equipment over time, only the operating model needs to be updated, the FMD software makes all its calculations based on it.

Model Software has developed this technology so it can be applied to systems with continuous-valued variables and has optimized the algorithms so that it can keep with a data rate of 1 sensor sample per second.

Intelligent Sentinel

The events of September 11, 2001 created an environment more aware of security risks. To help meet the challenge of protecting the nation, Model Software developed Intelligent Sentinel, a prototype highly advanced video surveillance system with patent-pending technology.

Intelligent Sentinel can track and record people as they move through watched areas. Also, the software is capable of tracking unique events, like unknown or hostile entities entering specific areas. Overall, Sentinel can be used as a force multiplier to give existing personnel a greater capability to protect an area.

More on Intelligent Sentinel can be found here.

Skills Demonstrated:

Model Software was called on to help correct and extend their existing web-based application. Previously, their software had been a source of much grief, but Model Software was able to help them correct the problems with their software and extend their product's functionality. provides logistics to the shipping industry.

Other Projects

Other projects Model Software has worked on