Machine learning that makes process control simpler

Conjecture System Identification Simple, best-fit models of your industrial systems for smoother process control

What if process control for industrial systems could be easier?

Conjecture System Identification uses machine learning techniques on industrial data to build models that reflect real-world processes - in a simple, intuitive way. These models can be used to improve control systems, reduce energy consumption, and improve the lifespan of existing instrumentation. Conjecture modelling can also aid in automating instrument maintenance, allowing for early detection of faults and maximising operation time.

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

The Benefits

local_gas_station

Better energy efficiency

Save on costs with enhanced automated control and select optimal operating modes

Improved reliability

Reduce moving parts and wear-and-tear, and stabilise your plant with amazing control

Behavioural diagnostics

Detect plant malfunctions early on by comparing model and underlying process

The Software

Conjecture System Identification is a software solution that models dynamic industrial processes by fitting best-fit trends to process data (such as SCADA data). The resulting model can aid engineers in enhancing the stability of their plant controller, reducing movement and wear-and-tear in their equipment, and improving the cost-efficiency of their system.

Conjecture System Identification is easy to learn and use with our online portal.
User tutorial videos are accessible from our Resources page, and email support is available.

Try online

User-friendly program interface

Enter and manipulate all your process dynamics in one place

Include delays and output shifts for estimation thanks to recent machine learning algorithms that run in the background

See your model equations take shape as you build - equations are in the time domain and easily understood

Specify initial guesses for your model estimates, or use Conjecture's defaults

See the results of your calculations in the Results pane - including estimated gain

Built in Plot Model function 

View your input, output, and model on the same plot

Zoom in and out to visualise how well your model fits the outputs from your process data

Watch how modelling cuts through the noise in your data, leaving the key patterns for observation

Stored results to help you choose the most useful model 

Conjecture stores every model you build during a session

Each model is stored with its results and ranked by goodness of fit (measured using the mean squared error)

Use this information along with your visualised plots to select the most useful model to characterise your system