Emerson Users Exchange meeting - Industrial plant computing on the cloud Making analytics less confusing Standardise to decarbonise
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Emerson Users Exchange meeting • Industrial plant computing on the cloud • Making analytics less confusing • Standardise to decarbonise Intelie - analytics for drilling Using unstructured data to reduce downtime May - June 2021 • Keeping hackers out of mobile apps • Working with vessel performance data • Blockchain for seismic, ESG, JVs and transport Official publication of Finding Petroleum
Operations pany involve safety, emissions and productiv- Depth and breadth and feedstock. Companies need to be sure they ity. But then the first step to achieving these have whatever they might need quickly avail- Providing such outcome based digital solu- goals is usually to achieve a certain level of able, but they do not have unlimited capital tions involves both depth and breadth. reliability with the operations. There will be and space available for goods in storage. But certain assets which are more critical than The depth element involves understanding the demand is often unpredictable. others in achieving a reliable safety level. Or how company’s businesses work, and how Baker Hughes has been working together with reliability may depend on certain aspects of new technologies work. For example, Baker the company C3.AI as its strategic partner, de- your supply chain. Hughes has projects to use AI on drill bit data, veloping AI based methods to plan inventory. enabling better insights into the condition of “Then you will be able to decide - what is the Mr Qasem’s business unit has worked with drill bits. It helps customers better understand missing part in all of this, in order to reach the C3.AI for its internal purposes, to reduce its how new technologies may be relevant to right outcome,” he says. inventory of stocks and to plan future manu- them, such as applying CT scanning devices Baker Hughes worked in this way with Petro- from the healthcare industry, or using 3D facturing, including making predictions of fu- bras. It did not start by saying “here’s a new printing. ture customer demand. system we’re trying to sell you”. It started off The breadth aspect involves connecting differ- Baker Hughes Digital Solutions with asking, “tell us the problem you face,” ent elements and understanding to get the right Mr Qasem says. Baker Hughes Digital Solutions is a multi-bil- outcome. Companies are seeking to break out Petrobras said it had over 6,000 sensor sys- of their old ‘silo’ structures and move to hav- lion-dollar business unit. tems across all of its assets, but had difficulty ing a horizontal digital enabled structure, with It provides a range of digital devices under a figuring out how the whole sites were con- better joining of the dots. Achieving breadth number of brands, including condition mon- necting together. Having this connected view can also mean deploying technologies at the itoring (Bently Nevada), sensing such as for was important for being able to answer ques- right scale, across the client’s entire oper- flow and temperature (Druck, Panametrics, tions such as, do we have enough feedstock. ations, not just in one place. Reuter-Stokes), control systems and cyber- Petrobras also had specific goals it wanted to Another example of breadth is where it helps security (Nexus Controls). Also industrial achieve, like “reduced greenhouse gas emis- a client tackle multiple goals at once. For inspection (Waygate Technologies, Process & sions”. example methane monitoring devices and Pipeline Services). Once these high level goals are defined, the systems can help with emissions, safety and It also provides monitoring systems, software next steps might be to install better ways to production at the same time. and services to connect all their data together measure flaring or methane emissions, and in a digital ecosystem. It serves 20 industries, connect this into a comprehensive digital eco- Inventory management not just oil and gas. system. “You will be able to find a better way An important part of achieving reliable oper- The company has its own R&D centre in to connect the dots,” Mr Qasem says. ations, and one which is often overlooked, is Oklahoma City, USA, which it calls better management of inventory of spare parts its “Energy Innovation Center”. Using unstructured data to reduce down time A lot of the insights which might help to reduce equipment downtime and improve maintenance planning are hidden in unstructured data in historical records. Subrat Nanda of ABS explains how this might be utilised By Subrat Nanda, Chief Data Scientist, ABS In the offshore industry, we are currently sat lates into costs of about $38 million. For the at the tip of an iceberg in regard to unlocking worst performers, figures are upwards of $88 the potential held inside unstructured data. million. If leveraged effectively, these data driven An effective asset management strategy can insights can inform decisions across an enor- start with combining data analytics with his- mous scope of critical business functions torical data and operational experience to re- – from improved operations and informed duce unplanned downtime and achieve higher planning to focused training and personnel operational availability. development. This involves fusing data generated from The potential to advance safety performance operations and prior maintenance, covering is also significant. diverse datasets from sources such as equip- Unplanned downtime, as we know, is costly ment design information, sensor time series for all offshore operators. A study by Baker data, maintenance records, inspection rec- Hughes found that 1% of unplanned down- ords, performance reports and class-survey time (i.e. 3.65 days a year) costs offshore oil reports. Subrat Nanda, Chief Data Scientist, ABS and gas organizations on average $5.037 mil- From this, an understanding of observed fail- Unlocking unstructured data lion annually. ure trends and risks can be gained, which in turn provides the data-driven insights needed One of the largest obstacles to obtaining The industry averages a little over 27 days maximum value from this exercise is that the of downtime every 12 months, which trans- to underpin condition based maintenance (CBM). maintenance history and operator observa- 12 digital energy journal - May - June 2021
Operations tional data is typically unstructured. form data to automate the task of identify- opposed to in siloes. This limits the achievement of CBM insights ing maintenance action types, differentiating maintenance scope and also isolating main- Condition based maintenance to be based mostly on structured parameter sensor data and in-situ or offline tests such as tenance triggers. Condition-based maintenance (CBM) is a vibration and oil quality. This results in faster, repeatable and more ac- maintenance strategy that dictates decisions curate analysis, leading to being able to iden- about what work needs to be carried out Part of my recent work has been looking at tify emergent issues, and model the reliability based on the actual condition of an asset. ways to better unlock the value of unstruc- tured data, usually stored in an operator’s risks in an asset fleet. Under CBM, maintenance should only be computerized maintenance management This formed a key part of our recent studies, performed when certain indicators are trig- system, repair and spare logs and other re- which involved developing advanced meth- gered and when it is economically optimal. positories. ods to perform natural language processing In other words, it should be done when there A typical offshore maintenance management customized to the unique problem domain of are signs of decreasing performance or up- system allows users to input free-text status marine machinery & equipment maintenance. coming failures. The maintenance should be reports. Many of the drop-down fields have We concentrated efforts on measuring, iden- performed at a time or location which makes missing or incomplete entries. tifying and improving data quality issues. it optimal from not only technical but also This involved building models to perform safety and economic perspectives. The problem with free-text fields is that they are written in a natural language format. automated annotation via various machine The need for CBM arises in part due to the How a person would speak, or using com- learning techniques. challenges of time-based maintenance prac- munity-specific terminology. This makes it a Several models using different hypothesis tices, as well as a need to reduce uncertainty major contributor to the poor quality of data spaces and relative strengths were tested for during maintenance events, and requirements generally found in a maintenance manage- building model ensembles. These included to safely extend the service life of equipment ment system. randomization-based methods, kernel-based to achieve maximum availability. Specific examples of data quality issues techniques, probabilistic models and in- Condition-based maintenance is not a re- include non-standard abbreviations used stance-based learning ideas. placement for subject matter experts (in by different operators and crew; inconsis- Following this work, we now have gener- fact, CBM relies on their input to train and tent equipment taxonomy; leaving critical alizable, accurate and automated modelling utilize their experiential knowledge to guide software fields blank due to lack of time or processes to extract insights from otherwise improvements). Rather, it is a methodology knowledge; common spelling and grammar unstructured and free text information, com- to inform maintenance strategies. mistakes; variation in sentence structures ing from the domain of diverse marine and Such information is only valuable if the data used to describe the same situation. offshore assets. fuelling CBM is of high quality – this is All of this means that datasets must be ana- In so doing, we have also developed a set of where data science comes in. lysed to extract useful information locked in artificial intelligence methods to address the Quality maintenance data (and thus improved an unstructured form before further analyses data quality challenge posed by unstructured analytics over time) can inform other funda- can be performed – a process which can be operations generated data sources. mental business operations such as human the difference between uncovering a systemic It facilitates faster and more reliable data pro- development and training. problem and letting it slip through the net. cessing and provides robustness against vari- For equipment manufacturers, it can high- ations in expression of semantics which are light common faults on the production line, AI for unstructured data commonplace in marine and offshore work- or identify equipment issues before they go Historical maintenance records have been ing environments. on to become widespread across the deployed used to train artificial intelligence algorithms fleet. It grants the ability to perform data fusion, capable of working with unstructured free- treating multiple sources of data as one as Digital.ai – keeping hackers out of mobile apps The apps which many oil and gas companies make, to provide their staff with sensor data, can also provide a pathway for hackers. Digital.ai has some tools to keep them out of mobile apps Digital.ai, a company based in Burlington, have no visibility of how it is being used, or ectly. The app code contains digital instruc- Massachusetts, has services and software ability to control or manage it. tions for how to communicate with corporate products to help oil and gas companies manage data systems, because someone would use the The apps are often developed by outside com- apps they build, to ensure that they cannot be app to engage with corporate systems. Once a panies, and can be widely distributed, includ- accessed by hackers. hacker can see the code, they can build a new ing to outside contractors. And the security app which can interfere with corporate systems It has a number of oil and gas industry cus- risks are often not well understood, says Paul in a malicious way. This is known as “reverse tomers, mainly with apps providing data from Dant, Vice President of Product Management, engineering” – based on taking apart a finished sensors. security with Digital.ai. product. Managing security of apps is a big challenge Reverse engineering For example, whoever designed the Stuxnet for oil and gas companies, because once soft- hacking software must have had access to the ware is in operation, the software creator can One of the biggest dangers with apps is that code, which was running Siemens turbines, in a hacker can open up the code and see it dir- May - June 2021 - digital energy journal 13
You can also read