Contributed by Kegan Smith
We are now in the Fourth Industrial Revolution (4IR) where real time technology service and solutions enable companies to translate acquired data into actionable results.
Mines and industrial processes in in remote areas have been updating to the latest software and digital sensors to move them into the Fourth Industrial Revolution, enabling them to monitor and control these facilities, remotely via internet communications and cloud data storage, which can even allow their venders to have access to this information, allowing them to supply spares, equipment and services ‘just in time’.
Most companies and software developers have focused on the equipment maintenance spares and expendables as well as the manufactured process product data.
Prei Instrumentation, industry 4 turbomachinery and setpoint (B&K Vibro) use high speed data collection and the power of OSI Pi to collect data from process machinery, and critical control loops to monitor and analyse critical process machinery. The level of expertise required to analyse this rich data is not always available on isolated sites and plants; by leveraging cloud services and connected partnerships Industry 4 turbomachinery can offer remote support and expertise in times of need or merely for routine reporting.
Rotating machinery is still at the heart of all major processes; whether it is compressors or turbines. However, currently, knowledgeable and experienced personnel for this critical skill is scarce and reaching retirement age. The control, maintenance and monitoring equipment and staff for rotating machinery is aged, and while the knowledge base is strong, much of the expertise remains in isolated departments with little unified migration to centralised intelligent databases; the industry is being left behind in comparison to the tools available today.
Industry 4.0 talks about a connected world; data streams coming from all corners of your environment with intelligence built into the technologies to make our life easier. We are seeing a growth in the smart sensor and the interfacing networks that connect the ground layer instrumentation. Sensors and the interfacing systems with built in ‘intelligence’ give rise to localised and self- diagnostics. This functionality is great for maintaining the health of the instrumentation and ideally suited for smaller repeatable balance of plant infrastructure. However, for large integrated machinery and process trains there is no context given to the diagnosis. Process instability can set off alarm bells pertaining to asset health indicators and vice versa, giving a false sense of diagnosis and leading you in the wrong direction.
The next concern in a developing industry is convention; parties talking about the same observations but from different perspectives, or worse, people talking about different observations but using common terminology. The concept of which is not our own, but fast becoming a means of levelling the playing field.
The importance of good clean data
The errors in data and its respective information content get compounded as one steps through the algorithms required to display, analyse and predict asset health.
Let us consider the flow of information. We start at the source… the measurement of the correct information, the reliability of the measurement and the trust in that information once it is converted into recorded data. The quality of the instrumentation is crucial to the trust placed in the protection, control loops and analytics alike. The same is to be said for the sampling of the information. Resolution in measured magnitude often surpasses the requirements, however poor time scales often lead to mistrust in the information.
Firstly, any engineer who has performed a root cause analysis on data depicted as quantified step trends knows the frustration of allocating a cause and effect. Next, remembering that we are discussing industry 4 and the analysis of data across systems; time synchronisation of the various data streams needs to be maintained at a similar resolution to the sampled data. Having a one second drift on data that is time stamped compared to a millisecond resolution is also problematic in assigning cause and effect (a problem that many are unaware of and think that the equipment is safe).
“Analysing and assessing rotating machinery such as compressor trains and blowers involves experience and the collection of the correct in-depth data.”
After measuring and sampling information comes the collection and storage. The high-resolution data and high sampling frequencies required for in depth analysis will saturate your data storage in short time if not managed correctly. There is always a level of filtering and compression involved in industrial historians. It is an old question that has plagued anyone with data and limited storage, what can I ignore without losing sight of the picture? Legacy systems rely on the crossing of finite thresholds and alarm points to trigger data collection with a small buffer prior to the event. The issue is a lack of information between events. Minor process hiccups and their increase in occurrence are often a key indicator to equipment degradation in health.
Another issue plaguing the Industry 4.0 development is the concept of data cycles. Data pertaining to a specific field of analytics or control is contained at the instrumentation and automation control layer with little or no transfer of rich data for statistical correlation across the process.
The modern architecture offered by real time data servers such as OSI PI allow us to collect data with a sub-millisecond time stamp from multiple data streams if necessary. In addition, we can implement custom compression algorithms based on the type of data. For example, vibration analysis waveforms can now be collected on parameter deviation outside a tuneable threshold.
Next in the journey of the data flow is the layers of analytics. Descriptive, versus diagnostic, versus predictive versus prescriptive.
Often systems that display data in a format specific to a type of data analysis, such as compressor map or vibration spectrums are described as diagnostic systems. However, these displays still require human intervention in the form of an expert to analyse and sift through the data. The diagnostics is not done by the system itself. This is the descriptive layer of analytics. The access to raw data and the ability to model that data into various display formats to highlight the information of interest in different states of failure and operation.
The next step is for the system to have algorithms which allow it to predict future degradation in the equipment based on the historical trend of the equipment condition. This still requires external interrogation on a frequent basis to interpret the systems prediction.
The last stage of the software capability is the prescriptive. How do we fold the thousands of tags that describe our assets into a one-dimensional health index of actionable data?
Remote collection and analysis
Remote collection of plant equipment condition data is now a reality with the Setpoint monitoring system.
The open architecture developed by OSIsoft is inherently secure. Setpoint as a condition monitoring system interfaces directly into the OSI PI environment. For plant wide installation the condition monitoring data can be collected in a local PI server correlated with process data and sent via PI can connect services. Industry 4 turbomachinery then retrieves this data to our office servers for remote support and reporting using tools as described through the data flow process. A similar architecture exists for isolated turbomachinery trains, where the data is sent directly from the Setpoint hardware to OSI’s cloud services.
These remote serves bring expertise to your door without lead times and expense of travel, therefore increasing support availability in situations of unexpected downtime.
Analysing and assessing rotating machinery such as compressor trains and blowers require experience and the collection of the correct in-depth data. The ability to correlate this data with existing information coming from the machines control and process information is crucial to the effectiveness and efficiency of such analysis. Distinguishing between cause and event relies on comparing these data sets for the same event.
Historical data should indicate signs and progression to the occurrence of such an event and should have the resolution in time and deviation to give early warning and possibly predictive action to mitigate catastrophic failure. At the very least, analysis of the data after an unfortunate catastrophic event should lead to root cause so that action may be taken to avoid reoccurrence.
The aim is to, cost effectively, use all the information at hand and the power of today’s high-speed computers, and the ability with the latest software developments to have these computers learning the correct responses and instructions required to avert any major catastrophic and expensive failures to remote, possible unattended, equipment. This form of artificial intelligence (AI) is now the way to go, however we must start with the correct data. Remember the adage: ‘rubbish in, rubbish out’.
About the author
Kegan Smith (31) qualified at University of Johannesburg (UJ) with Master’s degree in Electrical Engineering. He spent two years lecturing at UJ. He has been working for Prei Instrumentation for the past four years as a rotating machinery engineer, specialising in condition monitoring and analysis, vibration analysis and anti-surge control.
The Fourth Industrial Revolution (4IR) is the fourth major industrial era since the initial Industrial Revolution of the 18th century. 4IR is described as a range of new technologies that are fusing the physical, digital and biological worlds, and impacting all disciplines, economies and industries.