Dr Jacques Ludik is an executive director and co-founder of CSense Systems, headquartered in South Africa. Apart from his COO and CTO responsibilities, he is focussed on business development in Australia and Asia. Ludik completed his tertiary education at the University of Stellenbosch (Ph.D. (Comp. Sc.); M.Sc. Cum Laude; Hons. B.Sc. Cum Laude B.Sc Cum Laude), in South Africa. Here he started his career in 1993 as a computer science lecturer and researcher. Some 25 of his papers on artificial and computational intelligence have been published in journals and international conferences in the US, Europe, Africa, Asia and Australia. Ludik has also written a book, Neural Networks and Psychopathology, published by Cambridge University Press (UK). At the end of 1997, he co-founded CSense Systems (previously known as Crusader Systems). CSense software is distributed in Australia by Metquip Systems . Ludik recently visited Australia, where he spoke with Denes Bolza, editor of Process & Control Engineering (PACE).
DENES BOLZA: What is CSense?
JACQUES LUDIK: CSense is rapid process troubleshooting software that enables process engineers, production managers, HMI users, automation and control professionals, and other production staff to rapidly pinpoint and understand causes of production problems – and then take appropriate action. CSense’s front-end component, CSense Wizard, makes it easy to access data from HMI platforms, facilitates a troubleshooting methodology and reduces variations in key performance indicators, resulting in a more stable and optimised process. The wizard generates a solution blueprint that can be customised in a design-time environment, called CSense Architect, for real-time deployment. The blueprint contains automated data preparation, data-driven and rules-based models that have been automatically constructed in CSense Wizard.
How did CSense’s brand name originate?
The software allows the user to “see sense” in the complexity of a production or manufacturing process, identifying the most important variables that cause process deviation – for example, from targeted throughput and product quality. Users quickly make sense from large volumes of process data, after which they have many options to deploy real-time process solutions.
Why is “seeing sense” necessary?
Process variation is expensive, and can be hard to explain. Most processes are complex and dynamic, and it is not simple to relate relevant process variables to process performance. Furthermore, variation can be difficult to reduce, whether by understanding how to modify the operational philosophy, diagnostic solutions or advanced control. Reducing process variance leads to large monetary savings. Historian data in general is poorly utilised compared to the value that can be extracted with model-based troubleshooting techniques.
Earlier you mentioned the term “rapid process troubleshooting”. What’s the key here?
This refers to the methodology and technology implemented by CSense to improve throughput, quality, energy and data quality. In today’s competitive and cost-conscious markets, rapid production problem diagnosis and cure aren’t a luxury. Ten years ago, a product’s selling price was based on its manufacturing cost plus a profit margin. Today, where competition pegs the selling price, the only option for maintaining a profit margin is to lower the manufacturing cost. Here, an important component is whether the manufacturing process runs smoothly or not.
Where does CSense fill the “functional” gap?
It provides direct access to data in historians, databases and text files. CSense also has full OPC (OLE for Process Control) client and server functionality, providing real-time communication with the plant base layer control system or any other system that needs to read or write data from and to CSense.
In terms of production accounting, at which level does CSense fit in?
CSense creates a data acquisition layer just above the manufacturing execution systems (MES) layer. It takes data from PLCs and SCADA systems, validates it and reconstructs where needed. Production accounting includes data estimation and forecasting, with auto-updating and correction as laboratory data becomes available, reliability indexing of all tags, and real-time materials balancing. The all-important next step is to troubleshoot using that information. Troubleshooting can optimise processes in terms of their own efficiencies as well as that of their impact on equipment – resulting in a better return on assets. Thereafter, the production data is fed into the production accounting layer. The overall effect of data acquisition and validation, and the troubleshooting process, is to use production data to mitigate operational risks (maintaining peak plant efficiency consistently) and asset risks (ensuring equipment efficiency and longevity). To achieve this requires the ability to understand in detail the complexities of multiphase processes.
How are solutions arrived at?
Data collected from a process is verified and validated, using tailor-made data pre-processing blocks. These contain modelling and data processing functions, which build a blueprint of the solution. The models explain the causes of process exceptions and desired or undesired process behaviour. At each step of the blueprint, CSense checks data quality and suitability before any processing is carried out. This ensures that estimations, decisions or control actions are made only when process information is reliable.
Can you walk me through the various stages of solving problems?
The first step entails problem definition and data acquisition. In conventional analysis, more than 80 percent of time can be wasted in data acquisition and preparation. In CSense, this can all be automated. After graphical analysis and visualisation of the prepared data, non-linear and rules models are automatically constructed. Using a knowledge extraction dashboard, the user can then undertake cause and leverage analysis and what-if scenario analysis, and quantify the interrelationships between input and output variables. Process rules are automatically extracted from the historical process data. CSense then does a benefit gap estimation, using non-linear optimisation technology to demonstrate to what extent a variation in a key performance indicator can be reduced. The next step is what we call “knowledge fusion”. Here, the newly-discovered process knowledge combines with operational knowledge in an “action object”, or solution, which is then deployed real-time. A solution can take the form of calculating new optimal setpoints and sending these to a PLC or DCS, providing the operator with decision support, doing on-line mass and energy balances, etc. This visual programming environment saves many months in solution development, and still allows the user to do lower level scripting and incorporate user-defined functionality if required.
How fast is the blueprint?
It is able to execute on sub-second cycles, handling time delays and including historical data, such as lab analysis results. This ensures consistent and accurate analysis of all process events.
What has been your latest product release?
Version 3.3. CSense 4.0 is due out early in 2007. It will not only provide better support for troubleshooting via its continuous wizard, but also introduces a specific wizard to support the troubleshooting of discrete and batch processes. In primary materials plants continuous, time-based techniques will suffice. However, at least half the market has discrete and/or batch processes, variables and problems. CSense 4.0 will also enable action objects to be embedded in software industrial platforms like Wonderware’s ArchestrA. For this reason, CSense Systems has joined Wonderware’s independent software vendor program.
How does CSense complement Wonderware platforms?
CSense converts real-time data captured by Wonderware’s Industrial InSQL into useful information, using it to identify and solve problems. Interestingly, SQL historian was originally developed in South Africa and, after its initial success, was bought by Wonderware in California. In Australia, CSense can greatly benefit the InSQL clients of Metquip Systems. CSense also supports access to other historians such as Osisoft PI and GE Fanuc’s Proficy historian, and can communicate with any OPC-compliant third party platform.
Where can we expect further improvements in CSense’s partnership with Wonderware?
Instead of only connecting to ArchestrA via OPC, CSense 4.0 will make it possible to export CSense objects, or combinations of objects, that represent online solutions, to ArchestrA. End-users benefit because CSense can be used as a development environment to create solutions. The merging of such troubleshooting capabilities with ArchestrA’s agile control and information infrastructure leads to an unprecedented level of production and performance management and optimisation. The newly-extracted knowledge from CSense can be combined with existing expert knowledge. One such example might be operational best practices that can be expressed as fuzzy rules. CSense Architect is used to develop a more comprehensive solution with multiple components and layers, providing a visual drag-and-drop environment from libraries for rapid solution development.
Can you name some typical CSense applications?
Projects range from defect troubleshooting, process optimisation and stabilisation, process plant control, chemistry predictors, soft sensors, and equipment condition health monitoring to data validation and production accounting. In fact anywhere where you’d want to determine the causes of variation in production KPIs, to increase throughput, yield or recovery, or to improve quality and energy efficiency. CSense can also be used to reduce or eliminate operator inconsistency, increase the lifespan of equipment, or provide online mass and energy balances for improved materials accounting and production scheduling.
Who uses CSense globally?
CSense is used by many multinational companies in Europe, North America, South America, Africa and Australia. Examples include BHP Billiton, Rio Tinto, Anglo American, Mittal Steel, De Beers, Umicore, Zinifex, SABMiller and Sara Lee. The Finnish mineral processing giant, Outokumpu, uses CSense technology, as do vendor-independent CSense certified companies such as Hatch, which has successfully implemented a number of CSense projects (for example furnace diagnostics systems) in North America, Africa and also recently in Europe.
Why should anyone choose CSense over similar competitive products?
CSense has not only been designed for rapid process troubleshooting and to fully support a rigorous troubleshooting methodology that complements Six Sigma, but is also unique in the way it addresses major gaps in the market for rapid process troubleshooting. These include going from data to knowledge-driven action by making use of knowledge from that data, automating data preparation, and going from manual operational improvement actions to automated on-line action objects. The software also allows for rapid, low resource-impact troubleshooting and uses the power of advanced exploratory data analysis techniques while keeping it usable on a process level. CSense’s knowledge extraction dashboard makes it possible to identify causes for process deviation in an intuitive way, after which the user can do a benefit estimation using optimisation on the process model. It further allows the consolidation of new data-driven knowledge with existing expert knowledge into a CSense action object that can be deployed in real-time. The CSense real-time platform is also very robust in terms of online communication, handling of timestamps and quality, customising solutions, and exporting solutions to third party platforms.
What do you see as being CSense’s biggest challenges?
Our biggest challenges are growing our international distributor and solution partner network in an effective way, building the right international partnerships on an OEM and solutions level, and remaining market-driven.
How do you rate your business performance this year, and what has influenced it?
Sales are increasing fast, with volumes climbing by triple digits a year. The number of OEM and solution business opportunities is also growing, due to the positive feedback and successful references that are being created in our international reseller and solution partner network.