Predictive maintenance: the wrong solution to the real problem in chemicals

In the chemicals industry, Like many others, there is considerable excitement about the potential of advanced predictive maintenance(PdM)approach. The promise of these new technologies is tempting. The use of machine-learning techniques to comb through historical performance and failure data, their goal is to tell operators when and how the component is likely to go wrong in the future with a high level of predictability. This should reduce the impact of equipment failure and the costs of efforts to prevent such a failure-by the transfer efficiency is very low, not plan the maintenance activities effectiveness of the plan.

At first glance, chemical plants seemingly ideal environment seats. A high level of automation and instrumentation, combined with a strict maintenance record-keeping, create the wealth of data the machine learning system. In addition, most of the plant’s efforts for stable work conditions, may make it easier to find patterns and trends. There is also a compelling business case of improved reliability. Overall equipment effectiveness(preservation)losses due to unexpected repairs that range from 3%to 5% of the entire industry.

Prediction of poor outcome

A closer look, however, the potential of PdM in the chemicals began to evaporate, for four main reasons.

  • Too little data. In a chemical plant, the prediction of failure is more difficult than it first appears. Unexpected downtime is usually concentrated in a few large events. This means that there are usually too few data points for the PdM system of learning.
  • Too little time. Even when it can create the model the ability to predict, they tend to work time span is too short can be used for chemicals manufacturing. The prediction of this part of the Will cannot be in two days or two weeks is useful of the truck or mechanical tools, but it is unlikely to help in a plant shutdown takes a few days and the maintenance team need a few months to plan the interventions and sources of spare parts.
  • Too little impact. The impact from the PdM are often very low, because the factory work of critical assets with a high degree of redundancy and a few single points of failure. If a pumping station unexpectedly, the operator can usually switch to the backup unit of the small impact generated.
  • Too bar the province. Finally, the focus is to reduce unexpected downtime will ignore the largest source of throughput loss in most plants. Turn off scheduled maintenance events caused by the preservation of the loss of 5%to 10%of the average twice as much as unplanned downtime.

Digital reliability

Do these challenges mean analysis provides little or no value, and strive to improve asset productivity in the chemicals sector? No. The industry is achieving considerable success with a series of numbers the reliability of the technology, many of which are cheaper and less complex to implement than the advanced seats. To take three prominent examples:


To improve the situation of monitoring, through better remote sensing can cut mean time to repair, and significantly reduce the impact of equipment failure. In a chemical plant, some of the key pump suffered repeated failures. There is no backup available for these units, this problem is an important source of production loss in the factory.

“We decided we can’t wait for the the plant reliability and the engineer determine the root cause, re-determine the pump specifications, and then purchase the replacement,”Plant Conservation Management told us. “Therefore, we focused on mitigating the impact of failure, rather than avoid them.”

The plant reliability team to install some new sensors on the pump and start to monitor their status online in real time, allowing them to detect impending failure, a few hours before it happened. By being able to maintenance personnel ready to intervene, such intervention measures to reduce mean time to repair on these pumps from 6. 5 to about 3 hours, cutting overall equipment efficiency losses almost in half and saves approximately $ 120,000 for each failure.

Intelligent capital expenditure decisions

Better data means better investment decisions, especially when it comes to the allocation of sustaining capital expenditure costs or avoid equipment failure by making the correct, risk capital expenditure an informed decision. Most of the chemical companies it is difficult to set an appropriate level of continuing capital expenditures, because they find it difficult to allocate the funds in a plurality of factories and for different asset types.

This issue is ultimately about ensuring that resources are used to their maximum potential—this is exactly the problem, zero-base budgeting(this is nasty)have been successfully resolved through internal discipline, to assess all expenses of the return on investment. The same technique applied to the capital investment assets, or”asset it nasty”, which combines the available historical data and local expertise to assess the potential impacts of replacing or not replacing the particular device. The new approach allows all devices to update the project compared to the use of the same criteria: efficiency.

In a Manager, is responsible for the running of a chemicals manufacturer in the capital planning process,”we used to have a formal process to capture and evaluate the maintenance of capital expenditure items, but we have no clear way to rank order the projects and a summary of the beam priority to those with immediate, visible impact. Now we can have a fact-based, data-driven discussion about the risks and trade-offs, which allows us to spend less overall—and management what we do spend more wisely.”

Root cause problem solving

Better data also means better root cause problem solving. This helps the company to prevent the recurrence of failure, improve their failure mode and effects analysis(analysis)of processes, and optimization of the preventive maintenance plan. Together, these actions address the key aspects of the reliability, reduce the impact of failure and cost, prevent them.

In a chemical plant, for example, failure of a critical device caused the operator to activate an emergency shut down three times in as many months. These closures are inconvenient enough—but when the site team attempted to restart the plant, they found that the sudden closure of the unit, leads to the accumulation of solid in the key Tube and pipe. Fixed issue causing long delays in start and significant losses.

To solve this problem, the company applied combined with conventional root cause problem solving and informed analysis of the technology. Analysis of the data to help them understand how and why solids are accumulated under emergency shutdown conditions. The problem is the fixed combination of enhanced monitoring and changes to the preventive maintenance plan. But data-driven insights also allows the plant to modify its emergency shutdown procedures to stop the plant safely without causing the solid matter. This change reduced after the start time of any kind of emergency shut-off of 90%.

The potential for digital reliability far beyond preventive maintenance. And chemicals company, we believe that these other digital method is both easier to implement and provide greater value. Height of instrument nature of most chemical production facilities means that many companies already have a rich, and largely untapped, source of data to support the digital reliability of the effort. For those plants is not the time to”sensor”: the better the data is the critical first step of the digital reliability of the journey.

About the author(s)

Wim Gysegom Is a partner in McKinsey & Company’s London office, where Sven Houthuys Is a partner, Joel Thibert Is a partner in the San Diego office.

Spread the love
Do NOT follow this link or you will be banned from the site!