Improved business performance by analysing history

摘要

In recent years energy costs have been rising, along with new legislation being introduced on industrial emissions and control of waste material (IPC & IPPC & the change in Landfill regulations) combined with reduced profit margins, is forcing companies to look at how they operate. Many companies have in the past spent large amounts of capital expenditure to either improve plant performance or indeed completely re-design and build a new plant. In the current situation where profit margins are being continuously reduced, such options are not commercially viable. This therefore means companies having to look at the way they operate, this includes purchasing and stock of raw materials, manufacturing methods, manning levels, and distribution of end product. Whilst savings can be made in all areas, the most significant improvements can be made by understanding the processes involved in producing the end product and the parameters affecting it, and the interaction of one parameter on the other. This all sounds to be common sense, however special techniques have to be employed in order to achieve this. Modern day plants be they in the pharmaceutical, chemical, food or mineral industries are becoming increasingly sophisticated by the use of greater amounts of instrumentation and the introduction of DCS and SCADA systems using latest generation control software, can create their own problems in understanding the plant functionality. Whilst much of the plant is automated the end product quality is often measured by an offline technique, which can lead to delays in implementing control strategies to return the product to its correct specification. In consequence out of specification material is produced, which must either be reworked or scrapped, which leads to higher energy usage, lower yields, and longer cycle times. The advanced control systems collect vast amounts of data regarding the plant operation, some of this data is used for control, but much is purely recorded, and not acted upon. Even with the most sophisticated control systems if the plant output is still manually evaluated, optimum plant running conditions cannot be achieved, and understanding the interaction of the various parameters is not possible. The methods used to optimise plant performance and achieve the desired results, have been achieved by the use of recently introduced analytical techniques which allow knowledge discovery from Historical Data. It is the norm for modern manufacturing processes to implement comprehensive control and monitoring systems. One of the benefits of these systems is that they are able to provide huge volumes of data in a timely and usable manner: process flows, pressures, temperatures, qualities, energy use etc. These parameters may well be logged on a minute-by-minute basis (or even more frequently), and databases containing tens, even hundreds, of thousands of records of operating conditions can be made available. These huge databases contain knowledge about process operations, about the reasons for variations in throughput, quality, energy use, emissions, plant failures and similar. Finding this knowledge can be a complex task however, because more often than not, many parameters contribute to the variations in performance, there are complex interactions; and data volumes are huge. Few process organisations analyse such data effectively, if at all! Much of the knowledge is regarded as the "black art" of the experienced process operator. For example, in-line particle size analysis allows real-time tracking of full particle size distributions; data mining is a technology that allows this hidden knowledge to be discovered. These techniques are described, along with the results of initial work at a Chemical Production Facility. The work has aimed to identify new opportunities for reducing energy costs, and improving output and yield, whilst still main-taining product quality.