The pursuit of operational excellence (OE) can provide prodigious production and business benefits, as operating companies such as Lanxess and BASF and software suppliers such as KBC and AspenTech can attest. However, processors striving to achieve OE face many challenges, according to a survey last year by KBC and technical event organizer IQPC.
The survey asked leaders of more than 100 companies in the refining and petrochemical sectors about their understanding of OE and response to it.
“The research uncovered three main industry trends that reflect a changing approach to OE, shaped by emerging technologies and an increasingly volatile and competitive landscape,” notes Shane Fitzsimmons, global practice executive, strategy and business excellence, KBC, Houston.
The first is that most respondents have high hopes for OE. Indeed, 79% felt “very confident” that an OE mindset will result in safer, more reliable and more profitable operations. “However, it is clear that there are barriers preventing businesses from achieving this. It is especially concerning that only 25% of those we surveyed felt that their company has a clear vision for OE. Factors identified as barriers to operational excellence include ‘initiative overload,’ comfort with the status quo and internal organizational barriers,” he adds.
Second is the need for a new approach to OE. Respondents expressed concern not only about internal factors but also about increased volatility, competition and regulation in the market. In addition, trends such as changing workforce demographics, cyber security and technological disruption are coming to the forefront.
Across the board, leaders reported that their companies were in the “planning but not yet prepared” stage of readiness for these factors, says Fitzsimmons.
Third is the role that big data, the cloud, the industrial internet of things (IIOT) and artificial intelligence (AI) are playing. Fitzsimmons admits being surprised that some scepticism remains regarding the extent to which technology will influence the industry in the next five years. On one hand, many respondents reported high levels of maturity in adopting technologies such as advanced process control and maintenance/asset integrity systems. On the other hand, 65% said their companies weren’t ready yet to adopt AI while 41% said the same for the IIOT.
He stresses that the chemical industry must pay attention to three fundamental points: management of health, safety and environmental risk; fostering deep technical understanding as well as treating process data as a shared corporate asset; and creating an operations-centered mindset (Figure 1).
As an example of this in action, Fitzsimmons cites an OE program that KBC developed and rolled out for an East Asian refiner that was experiencing issues with loss prevention, operating efficiency and aligning operating practices with the wider corporate plan.
The strategy focused on four major themes over three years, he notes. “Firstly, we worked to conceptualize and design the OE model, in consultation with senior leadership. The second pillar focused on margin capture. By implementing operational improvement, the refiner ensured increased short-term profitability, turnaround optimization and curtailment of potential production losses. Thirdly, KBC recommended a major realignment and focus on developing basic capability in the organization. Finally, we took steps to address fundamental safety behaviors and risk management.”
Fitzsimmons points to impressive results from implementing these steps: The refiner executed a turnaround requiring 20 million manhours without a lost time incident. Financially, it achieved $160 million in cumulative benefits, with a further $250 million in potential benefits identified.
“The number of truly excellent organizations is small, with increasing divergence between typical and world-class operators. By applying emerging digital technology and applications, guided by robust business processes, businesses in the chemical and refining industries can achieve superior, sustained results,” he emphasizes.
Lanxess, Leverkusen, Germany, is using its manufacturing excellence (MEx) initiative to hone OE. “One goal of MEx is to operate plants at optimal costs. The main target here is to optimize production processes in order to have a high availability and a high yield of the plant and so achieve maximum productivity. In addition, our team draws a special focus in the value stream analysis on product losses via gas, liquid and solid waste streams. And the positive effect is twofold: less material consumption for the production process on the one hand and reduced waste treatment cost on the other,” explains Ralf Krueger, head of MEx.
He cites as a typical example a batch product line in which the company instituted a variety of measures to free up significantly more of the available production capacity: “This enabled the insourcing of production that had been done by a toller before. The effects were cost savings and also a more flexible production schedule, which led to higher customer satisfaction.”
Lanxess, which acquired Chemtura in 2016, now is extending MEx activities to former Chemtura plants as part of the overall integration process. “The projects were a good platform for getting to know each other’s cultural background better especially at site level. Moreover, they allowed sharing of best practices and enabled knowledge transfer. By working together, it was very easy to cross-link the new sites with all different disciplines and competence centers Lanxess-wide,” adds Krueger.
Another important part of the MEx process is digitalization, especially in terms of supply chain excellence. To this end, Lanxess has transferred the activities of its commercial and supply chain excellence initiative to a newly formed group responsible for digital transformation.
“As the MEx initiative is a great success, the program will continue as a permanent entity. There are two main objectives going forward: to identify improvement potentials in the mid-size plants which have not yet been part of the program and to act as a hands-on catalyst in the integration process of new acquisitions,” he stresses.
Meanwhile, the ongoing strategic OE program at BASF, Ludwigshafen, Germany, called DrivE, is on course to deliver the anticipated €1 billion ($1.14 billion) in earnings every year compared with a 2015 baseline. “OE is a decisive factor for our competitiveness and a driver for sustainability,” notes a spokeswoman.
Many DrivE projects rely on a comprehensive methodology known as technical process optimization (TPO) to help improve productivity. “This focuses on increasing plant reliability, reducing raw material and energy consumption and expanding our production output, where needed by the markets. Experienced process engineers team up with all required subject matter experts and the plant teams to jointly analyze the current operations and develop concrete measures to improve productivity,” says the spokeswoman.
A dedicated team within BASF has used TPO for 15 years now, optimizing more than 500 production plants along the way. Debottlenecking often provides capacity benefits and usally affords savings in both fixed and variable costs. Such OE measures typically boast a payback of less than a year.
Two of the most recent TPO projects have targeted the company’s Care Chemical division, which operates plants in Antwerp, Belgium, and Freeport, Texas. Both manufacture superabsorbent polymers used in hygiene products. “These 2017 projects yielded reliability improvements, energy savings and product quality enhancements that have overall achieved significant savings in costs,” she explains.
In addition, BASF — as part of its predictive maintenance strategy, which also falls under the DrivE OE umbrella — implemented projects in five plants through the end of 2017. The plan now is to execute predictive maintenance projects at a further 34 plants worldwide by the end of the first quarter of 2019.
Meanwhile, at AspenTech, Bedford, Mass., the OE focus is very much on hunting down signatures to equipment failure.
“Process disruptions lead to lower production levels, off-spec products and potential failures. Low asset availability and unplanned downtime result in lost production and excessive maintenance costs. Low asset availability increases capital costs and lowers production capacity. The solution is deep process domain and modeling expertise combined with data science to provide the rich context necessary to accurately predict and avoid asset performance issues,” says Mike Brooks, senior director, asset performance management (APM).
Aspen’s main tool here is aspenONE APM and its related products Mtell, Fidelis, ProMV and Column Analytics. APM uses asset-specific analytics to provide early detection and guidance to help operators keep the process away from conditions that are detrimental to the asset. Reliability modeling supplies the information needed to make economically optimal decisions and to focus on the most vulnerable assets or failure modes, while machine learning agents scour real-time data streams to predict situations before they cause losses or asset damage.
“We do not model machines, we measure signatures of failure, Brooks explains. “For example, Aspen Mtell uses machine learning to identify multivariate, temporal patterns that humans cannot see and they become the core signatures of agents — small pieces of software that do the work so the end-user does not have to. Agents recognize normal and abnormal behavior of assets and the specific patterns of degradation that lead to failure.”
Unlike conventional engineering/mathematical models, agents are based on the actual, real-world behavior of the equipment (Figure 2), he stresses. They take into account all conditions: seasonal, different operating campaigns and duty cycles as well as deterioration of process and mechanical performance.
“The technology deployment model behind agents — it’s important to say it this way, because it does not come inherently just from machine learning — does not care about what type of machine, the industry vertical where it is used, or the engineering principles behind its operation. It only cares that there are sufficient sensors supplying enough data that contain learnable relationships between the sensors to accurately declare the operating behavior of the asset through normal and degradation/failure circumstances.”
In this context, Brooks counsels processors to focus on getting answers to several crucial questions: Does the proposed option actually provide better — i.e., earlier or more accurate — information? Is it fast and easy to do, with little training? Does it close the competency gap, so existing plant staff can do the work, both in terms of implementing the technology in the plant and in making sense of and acting on the information it provides? Does it address the number one issue causing degradation and breakdown? Does it get better over time? Is it scalable across a single plant, a site, and a whole corporation?
He cites a number of examples of how this approach has worked — and also the challenges it has faced.
A refiner couldn’t solve its compressor problems with state-of-the-art vibration systems or reliability-centered maintenance techniques. So, it turned to Aspen Mtell. The machine-learning software identified the root cause of the problem causing equipment breakdowns. However, the operations staff chose not to heed the warning. Seven weeks later, vibration became so intense that it necessitated a three-day shutdown. During this period, personnel disassembled the compressor and found the issue to be the exact one predicted by the software seven weeks earlier. Process changes made since then have avoided any further damage.
A very similar situation occurred at a large North American oil and gas producer and refiner. On the third day following implementation of Aspen Mtell to protect three major compressors and pumps, one anomaly agent issued an alert about a compressor problem. Moreover, it exposed the root cause of this problem, which had plagued the refinery for more than a decade. Again, operations staff chose to ignore the heads-up; an urgent shutdown took place eight weeks later. Investigations revealed that Aspen Mtell had correctly warned of the impeding failure.
Early warning about looming breakdowns can foster prescriptive maintenance, as a major pharmaceutical plant can attest. It was suffering continual failures in its chillers and compressors that caused high production losses. Now, the plant relies on Mtell’s agents for guidance on when to maintain equipment. The software gives sufficient notice to enable orderly and rapid problem correction at the lowest cost. Moreover, it has led to dramatic improvements in overall production valued at millions of dollars per year.
In February, AspenTech formed an alliance with Emerson, Austin, Texas, to further help customers optimize production and drive OE. “It’s a very important relationship and allows us to combine our technology with Emerson’s remote inspection tools to get even earlier warnings,” notes Brooks.
Seán Ottewell is Chemical Processing's Editor at Large. You can email him at email@example.com.