This icon is a registered trademark of GoalArt. Click to go to the GoalArt home page.
  The name 'GoalArt' is a registered trademark. Click to go to the GoalArt home page.
  GoalArt offers unique methods for describing process knowledge and using it in automated diagnosis. GoalArt's methods provides situational knowledge for the operators. In this way, a GoalArt system provides a unique degree of safety for a system. This is why 'Knowledge for Safe Operation' is our motto. Click to go to the GoalArt home page.
On-Line Demos
News and Publications
Products and Services
Industrial Branches
Diagnostic Functions
Analytic Services
Technical Background
An Alarm Analysis Example
Other Techniques
GoalArt Diagnostic Station
Software Licenses
Feasibility Studies
Modeling Projects
Training Courses
Service and Support
Customers and Projects
Service and Support
Career Opportunities
About GoalArt
Contact Us
Site Map

GoalArt offers a unique set of methods for advanced decision support. Our methods can solve the alarm problems of industrial plants and complex technical products. Our technology is superior to older techniques.

Here is a short comparison with competing but older techniques.

There is no serious competitor to GoalArt's technology today.

Current systems offer no support for tasks such as alarm analysis, fault diagnosis, and on-line what-if scenario analysis. Suppliers may offer to "solve" your alarm problems, but what you get is an expensive tuning effort, which, at best, only removes nuiscance alarms. The alarms in a alarm cascade are not wrongly tuned, they are caused by other alarms. Only a method for separation of primary and consequential alarms will sort this out.

Current systems may offer the possibility to assign static priority levels to alarms, so that more important faults can be treated first. GoalArt offers state-based alarm priority, which allows for dynamic priority levels depending on the current operating state. It also allows you to suppress alarms in states where they are not relevant.

Static alarm priority cannot solve the problems of alarm cascades. The most important fault may be a consequence of a less important one. Instead of treating the important fault, it may be possible to alleviate the whole situation by a simple action on the root cause. Only alarm analysis can help you with this.

Alarm analysis can be performed using fault trees or control system logic. Here, you have to take care of every fault combination and verify that the fault trees are correct, which means that it is impossible to produce a complete solution. Even a small system with 15 sensors (low / normal / high) produces more than 50 000 000 cases.

Thus, a fault tree solution is always incomplete and unreliable, while GoalArt's alarm analysis will handle all theoretically possible cases correctly. Building an MFM model is several orders of magnitude quicker, and MFM covers everything that can be expressed in fault trees, so fault trees have no advantage over MFM.

You cannot solve these problems by lot's of programming, no matter how much. You need methods solidly based in theory.

When it comes to technical products such as airplanes, cars, and medical equipment, any fault diagnosis is usually handled by writing dedicated code. GoalArt's methods demand much less effort in construction and especially in updating, and they have excellent properties for integration in standard code. GoalArt's algorithms demand very little CPU power and memory, and they are written in standard C and C++.

Some suppliers use rule-based systems with either crisp of fuzzy logic. However, the knowledge engineering effort needed to build a rule-based system is much larger than with our methods, and the real-time properties of the resulting systems are often doubtful. Finally, it is difficult to verify the correctness of a rule base. In short, a rule-based system suffers from the same drawbacks as fault trees and alarm logic.

Diagnostic methods within artificial intelligence are often based on predicate logic and use very general reasoning mechanisms, such as, for example, Reiter's algorithm. In research, this is sometimes described as the solution to fault diagnosis. In fact, these systems have been used mainly in the domain of testing electronic circuits and not in other areas. The methods are usually computationally complex and unable to handle more than a few hundred inputs. Finally, they are still research, not products.

In research, there is a strong interest in learning methods, based on, for example, Bayesian methods, belief networks and neural networks. Instead of going through the hard work of model building and knowledge engineering, why not let the system learn from experience? But where do we learn to see the patterns of unexpected accidents, or once-in-a-lifetime events? GoalArt's methods use available design knowledge to understand situations that may never have occurred before, even once. It is a drawback to have to learn when you can get it correct from the first moment, and a learning system will always be unpredictable when faced with new situations.

Today's control systems have a much better presentation that yesterday's systems. With easy navigation between nice screen pictures, operators can understand even complex fault situations. Correct? No, this is not true. No presentation, however nice or understandable, allows operators to understand large alarm cascades alarms, like the one that occurred in Harrisburg. The only way of handling alarm cascades is to have an automated root cause analysis algorithm help the humans.

In conclusion, GoalArt offers a set of methods for advanced decision support, and these methods are available now. We are strongly involved in research in artificial intelligence and automatic control, but our methods are mature and available, and we are ready to deliver now. We offer products, not research projects.

GoalArt, Scheelevägen 17, 223 63 Lund, Sweden, Phone: +46 46 286 4880, Fax: +46 46 286 4882, E-mail:, Web: