Conjecture is a system identification software solution which allows you to create best-fit models of your industrial process data (such as from SCADA) using machine learning techniques, to better understand, diagnose, and control your industrial systems.

System identification is the method of using mathematical models to characterise dynamic systems from measured data. Dynamic systems have measurements of inputs and outputs to the system that change with time.

Conjecture can be used in any industry which requires automated feedback loop control of processes, including oil and gas, minerals, and manufacturing.

The Conjecture System Identification software solution was developed by Robert A. Anderson, an electrical and process control engineer from Perth, Western Australia. In 2011, Robert began exploring the mathematics of system identification in process control. He realised that the models and methods commonly taught at university for system identification were difficult to implement in everyday engineering work. Furthermore, currently available system identification programs did not accommodate important process dynamics such as delays and output shifts. This resulted in models that were a poor representation of real-world industrial systems.

Over several years of independent work, Robert identified the machine learning techniques that could be used to model industrial plant dynamics straight from the data, incorporating both delays and output shifts. He then set about creating a simple software program to apply these techniques that would be intuitive to fellow engineers, finally completing the first prototype of Conjecture in 2017.

Most system identification programs are based on the theoretical models commonly taught in engineering study, such as z-domains and Laplace transforms, which are difficult to use in daily engineering work. In contrast, Conjecture helps the user to specify a simple model for their data based on time series equations – X’s and Y’s with time delays and an output shift. It then runs all the calculations in the background, producing easy-to-understand results that translate directly to the real-world system.

Goodness of Fit
Conjecture is the only system identification program readily available on the market that allows users to model both delay and output shift in their systems. This produces models that are a much better reflection of these systems. All these dynamics can be modelled and understood easily in the Conjecture interface.

By adding an output shift with the click of a button, Conjecture modelling allows the user to estimate the output level corresponding to a zero input. This capability is essential to producing a useful model, but is not easily accessible from some other system identification programs.

Furthermore, Conjecture allows users to estimate the delay in their system. Delay is the time lag between a change in input and the corresponding change in output. Industrial systems with delay are difficult to control effectively since traditional control algorithms cannot compensate for this time lag. By estimating delay, Conjecture gives engineers the information they need for smoother, more stable control.

Conjecture has a graphing function that allows users to view their input data, output data, and model on the same plot, to easily assess how well the fitted model reflects their real-world data. This assessment is crucial to choosing a useful model.

Through intuitive best-fit modelling of your plant data, Conjecture can benefit your plant operations through improvements in:

By modelling your plant’s typical operation using Conjecture, you can more easily identify periods of atypical operation where the data produced by your plant deviates from the model. Problems such as valve passing or blocked pipes could be spotted early on, before they cause significant interruption to your operation.

Conjecture model provides greater understanding of the dynamics of your plant. You can use this information to train operators and engineers, and even to experiment with the operation of your plant in the digital world – where experimentation is safer and cheaper.

Most process engineers would say that their biggest challenge in controlling an industrial system is accounting for time delay. Model-fitting with Conjecture allows you to quantify the delay in your system, giving you the tools to control your system more stably. The Conjecture user training videos include a demonstration on how to do this.

No – with just a few quick training videos, users can adeptly load and model their data in Conjecture. The simplicity of Conjecture is what sets it apart from other system identification software.

No. Conjecture uses different mathematical techniques compared to traditional system identification programs – techniques that are easier to grasp and more intuitive than the z-domains and Laplace transforms usually taught in engineering study. Also, most of this mathematics runs in the background, behind our simple, easy-to-use interface. Most users can adeptly load and model their data in Conjecture after a short training session.

Conjecture has been designed to work on systems with a single output and one or more inputs. You can try modelling you data using our online portal. Feel free to contact us if you have further questions about whether Conjecture will work for your system.

We have just released our online version of Conjecture here. This will soon have facilities enabling you to log-in and subscribe. If you have any questions or need any assistance using the portal, please contact us.

At present, the Conjecture System Identification solution stands alone as the flagship product of Conjecture Data Solutions. It is the result of years of development and represents an approach to data that we believe is the future of the engineering industry. Hence, we are focusing our efforts on establishing Conjecture as the process control analysis software of choice in Western Australian engineering. In the coming years, we will be looking to develop other software products in response to the needs of the engineering community.