Brendan Thorne

Aughinish Alumina Limited


AAL have implemented a wide range of sophisticated applications covering process control, information, and scheduling systems. The replacement of the Taylor Mod 3 with a Honeywell TDC 3000 Distributed Control System (DCS) between 1990 and 1993 provided the platform that facilitated these developments. The installation of PI (from Oil Systems Incorporated) and later the G2 Expert System (from Gensym Corporation) were part of this program. The overall system played a key role in the expansion of plant capacity from its original design of 800,000 tpa to its current capacity of 1,430,000 tpa.

The paper details the background to this program, outlines the design and implementation of the individual applications, and quantifies the resulting benefits.

The most significant areas of advancement have been:

Areas where current development is taking place are:


Brendan Thorne


Aughinish Alumina Limited (AAL) has achieved significant benefits from its investment in automation, control, and information management. The base platform for this was the installation of a Honeywell TDC 3000 control system between 1990 and 1993. A PI system was installed at the same time, giving all personnel access to process and other data. From 1997, Gensym's G2 expert system has provided a supervisory control facility. The improvements came from applications developed on all systems.

The main benefit has been increased production, but the systems have also led to improved efficiency and quality and facilitated manpower reductions.


The AAL plant was designed in the early 1970s and had a nameplate capacity of 800,000 tpa. The economic uncertainties following the 1973 oil crisis dictated that construction did not begin until 1978, and the plant went into production in September 1983. Construction followed the original design, and this meant that the control system, a Taylor MOD 3 with associated Taylor 3106 and 3103 process computers, was obsolescent on start-up. In September 1987 Taylor informed AAL that the system was effectively obsolete, and that only limited cover would be available from 1988.

Alcan and Billiton, the partners who owned AAL at that time, decided in 1988 to proceed with the installation of a state of the art Distributed Control System (DCS). The main justification for the project was replacement of obsolete equipment, but there was an expectation that the new DCS would yield significant improvements in throughput and process efficiencies. The selected vendor was Honeywell, and the system was installed between February 1990 and June 1993. An information system was required to replace the Taylor 3106 Process Computer, which generated data for the daily operating report and motor run log. As part of the DCS Project, AAL installed the Plant Information (PI) system in 1990. The vendor was Oil Systems Incorporated, now known as OSI Software. In 1996 AAL installed the G2 Expert System. The vendor is Gensym Corporation.


AAL's TDC 3000 system is based on Honeywell's Process Managers (PMs). They provide facilities for analog and digital input and output, standard control algorithms, digital control (on/off operations), logic, and sequence control. Each process area has a set of PMs operating on its own network. AAL has 5 such networks, covering Digestion, Mud Circuit & Filtration, Precipitation, Calcination, and Boilerhouse. The PMs on each network can communicate with each other, but not with other networks. Inter-network control is done using Honeywell's Application Module (AM). Over the years AAL has upgraded to the Advanced and later the High Performance Process Managers.

The PI system was originally installed on a VAX 6310 computer, and accessed through high-resolution Tektronix monitors. As the use of personal computers grew, PI was increasingly viewed via X-Windows, but in its original format. From 1994 OSI developed PI Process Book, a windows compliant package, and also introduced the PI Excel link, allowing direct access from spreadsheets to the data archive. An Alpha 4000 has now replaced the original VAX 6310.

G2 was installed on an Alpha server in 1996.


4.1 Overview of Benefits

The main benefit that has resulted from installing the control and information systems is increased production. It is difficult to quantify the financial implications of these systems precisely, but a 1992 study identified $2,040,000 per annum in cost savings and increased production directly attributable to the DCS and PI systems. The value has undoubtedly increased since then, but has not been quantified.

The introduction of the DCS allowed AAL to merge two control rooms into a single control room for the mud circuit in 1992. This was repeated when boilerhouse and calcination control rooms were merged in 1994. In both cases a single Control Room Operator (CRO) controls the process.

Efficiency and quality have also benefited from the introduction of the more advanced control schemes. The PI system enabled AAL to reduce manpower, and was significant in facilitating a major reorganization introduced in 1993/94.

Individual applications are now considered.

4.2 Plant Information (PI) System

Although PI was installed to replace the Taylor 3106 Process Computer, it was not long before all plant personnel realized how direct access to process and other data could improve plant operations. The initial estimate, considered wildly extravagant at the time, was that no more than 15 persons would want to view PI. Typically, about 150 users now log on to view the process through PI.

The original system, Classic PI as it is now called, was mainly a trending facility. AAL policy was to build tags to allow the process be viewed as close as possible to on-line. Thus all tags update every minute, if necessary. All analog and digital inputs on the DCS are replicated on PI, and in the case of controllers the setpoint, output, mode, and tuning constants are also built. The main benefit is the ability to troubleshoot existing problems and diagnose the cause of process upsets after they have occurred. In 1993 OSI developed a graphics package, and this allows the plant status to be represented on a single screen, with subsidiary screens to follow individual operations. In 1994 OSI introduced PI Process Book. This is now fully compliant with Windows, and the most recent version includes the facility to use Visual Basic to enhance displays. In 1994 OSI also brought out the PI Excel link, allowing spreadsheets direct access to the PI archive.

It is difficult to quantify the benefits of the increased access to information. The fact that the direct cause of process problems can be identified quickly means that they do not last as long, and this increases production and improves efficiency and quality. PI is also used to analyze plant performance over time and identify the reasons why the performance changes from one period to the next. Finally, PI is now the basis for reports that formerly required extensive manual effort to get the data and verify it, thus reducing clerical effort.

Although the majority of users continue to use Classic PI, AAL plan to standardize the use of Process Book in the coming year. This will provide a single format in which to view process data, access reports, manuals and maintenance information, write logs and get any other relevant information.

4.3 Sand Filter Optimization

AAL uses 16 Alcoa sand filters to remove the last traces of particulate matter before the pregnant liquor enters the precipitation circuit. The filters are discontinuous in operation, requiring periodic backwashing to remove accumulated mud. Prior to the DCS installation, the sand filter building had a dedicated CRO and Field Operator. The process was semi-automated, with a Taylor 3103 Process Computer sequencing the operation of actuated valves and the MOD 3 system controlling flows. No communication existed between the two systems. Backwashing required the CRO to operate the two systems and communicate with the Field Operator. This led to the sand filters being the most significant bottleneck in the plant, with a maximum of 24 backwashes possible per day, and limited plant flow to 1,800 m3/hr.

When the DCS was installed, each filter had an identical, though separate, sequence program controlling all aspects of its operation. On completion of the project, the sand filter control room was merged with the mud circuit control room without any problems, although the Field Operator remained in the filter building to operate the non-actuated valves. In 1993 these valves were actuated and incorporated into the programs, and the Field Operator reassigned to other duties. A system was installed on DCS to allow the CRO to preselect the filters for backwashing on his shift, and the sequencing then occurred without any external intervention. This system allowed an increase in the number of backwashes per day from 24 to an average of 55. The elimination of the sand filters as the main plant bottleneck allowed plant flow to be increased from 1,800 to 2,300 m3/hr. The installation of the DCS and the upgrading of valves from manual to actuated thus increased production by 28%, reduced manpower per shift by 2 (for a total of 8 with the 4 cycle shift), and consolidated two control rooms into one.

4.4 Bauxite Charge Control

AAL uses high temperature digesters to extract the boehmite in the bauxite. The original control procedure was completely manual. Lab samples were taken at the blow-off tank and, based on the results, the Control Room Operator calculated what change in bauxite slurry flow was required to achieve the target. In 1990 an on-line measurement system was installed. This measured the density, conductivity, and temperature of the slurry and used these parameters to calculate the caustic, alumina, and ratio of the liquor. The large deadtime and lag between the addition of bauxite and measurement of the DBO ratio meant that a conventional PID loop could not be used for control.

In 1992, AAL engaged the services of Icotron, a control consultancy firm wholly owned by Honeywell. They designed a model based control scheme, using a Modified Smith Predictor for deadtime compensation. The model assumes that DBO ratio is a function only of the bauxite slurry/plant flow ratio. Two controllers dictate the changes in bauxite slurry flow. The analyzer controller takes its input from the on-line calculation and is a standard PI controller. The Modified Smith Predictor program calculates the predictor controller's input; this is an integral only controller. Bauxite slurry flow is set as a ratio of plant flow, and the setpoint of this ratio controller is dictated by the outputs from the analyzer and predictor through an incremental summer. Their tuning constants are calculated from the process gain, deadtime, and lag. The scheme can operate using on-line and Lab data as inputs, but the analyzer's integral action is turned off when using Lab data.

The scheme was an immediate success, and reduced the standard deviation on DBO ratio control from 0.008 to 0.003.


4.5 Optimization of DBO Ratio

The sand filters used at AAL limit the maximum DBO ratio because high ratios cause cementation within the filters. Experience has shown that we can operate safely if the ratio of the liquor entering the filters is not supersaturated above a margin. A model was developed to predict the temperature entering the sand filter building knowing the plant operating conditions, and from this we calculate the approach to supersaturation every 15 minutes on the DCS. If the current DBO ratio target contravenes this margin the CRO gets an advisory alarm, and he adjusts the target if he believes that the process conditions justify him doing so.

The DBO ratio target may also be limited by conditions in the digester. We calculate the digester caustic and ratio through mass balances. The caustic and temperature are used to calculate the equilibrium digester ratio, which is the ratio expected if ideal conditions applied. The digester margin, i.e. the difference between the equilibrium ratio and that in the digester, is the main control parameter. When a DBO sample is taken, the DCS records the timestamp. When the Lab download the boehmite result the DCS automatically informs the CRO that a new result is available and also calculates the conditions in the digesters when the sample was passing through. The most significant operating parameters are the digester margin, as mentioned above, plus the residence time and lime charge. Knowing these, the CRO can pinpoint the cause of poor extraction and take appropriate action.

4.6 Effluent Management

AAL has two main sources of effluent. The first is process condensate, which has little soda but is hot. This is neutralized in a single step process using fresh (20%) sulphuric acid and generates comparatively little solids. The neutralized stream goes to the Liquid Waste Pond where the solids settle, the condensate cools, and from here it goes to the river Shannon.

The second source of effluent is from the various plant ponds. The Storm Water Pond takes the runoff from the mud stack, and the other ponds collect runoff from the plant and dump streams. Depending on the flow and degree of contamination of each stream and the plant volume situation, they are either returned to the process via the mud wash circuit or collected in the South Pond. The South Pond is neutralized, but unlike the condensate, this generates significant amounts of solids. Experience has shown that best results are achieved by neutralizing in two successive buffer tanks and settling the resultant slurry in a clarifier. Spent acid from heater cleaning is added to the first buffer tank, this being the optimum means of disposal.

It is impractical to manage all these streams using conventional control techniques, particularly as the component streams are not controlled. As a first step, the G2 expert system was used to control the return flow from the South Pond. It also manages the spent acid disposal, making gradual adjustments in the flow depending on the availability of spent acid, its use elsewhere in the plant, and the neutralization requirement in the first buffer tank. These modifications have greatly improved the control of the effluent system and thereby increased its capacity.

In 1999 a second clarifier will be installed. This will include facilities to measure the flow and conductivity of each stream. G2 will then direct the streams to their optimum destination, depending on their degree of contamination and the plant volume situation.

4.7 Calciner Fuel Optimization

AAL operate three Alcoa fluid-flash calciners. Over the years, modifications have increased their capacity to 5,000 kg/hr fuel rate, which corresponds to 63 t/hr alumina. However, the CROs are not at ease operating at this rate because deviations from normal conditions will lead to a trip. To overcome this, AAL developed a G2 application that reviews 67 different operating parameters for each calciner. If all are within their normal range, G2 will progressively increase the fuel rate up to an agreed constraint limit. Deviations from normal will lead to reduced fuel rate. The logic used is fuzzy, with appropriate membership functions for each parameter. The CRO has full responsibility for the decision to apply the optimizer, and also to select the constraint limits. The initial application sought to advise the CRO, but the feedback was that they did not like getting suggestions about what to do, and preferred to close the loop completely. The application has been on-line since early 1997 and has given an average of 5% increase in calciner throughput.

4.8 Specific Surface Area Control

The specific surface area of the calcined alumina (BET) is an important quality parameter. Although specific surface area is known to be a function of holding vessel temperature and residence time, early attempts to control it on this basis were not successful. In 1997 AAL implemented a G2 scheme to resolve this problem. The Lab BET result is checked through fuzzy logic that evaluates how it is positioned with respect to the control band. The overlapping membership functions mean that the resultant value is not simply inside or outside the band. The result is converted into a change in the holding vessel temperature target, and downloaded to DCS.

When the calciners operate at lower fuel rates, recirculation will affect the BET result. To overcome this, separate logic adjusts the temperature target of the main furnace as required.

At the same time as these control changes, mechanical modifications were made to the holding vessel. It is impossible to assign cost benefits between the two, but the overall impact has been to reduce the standard deviation on BET from 10 to 6.

4.9 Agglomeration Control

In 1996 AAL installed BTG probes to measure and control the fine seed charge to the agglomeration section of the continuous precipitation chains. This was successful, but the true control parameter here is not gpl solids but rather the surface area in the seed charge. This cannot be measured on-line, and the CRO had to adjust the seed charge based on the Lab results. In 1998 AAL developed a G2 application to do this automatically. The scheme has two parts.

The feedforward part looks at the change in the main parameters such as temperature, caustic, and ratio in the agglomerator, -45m fraction of the fine seed, and crystal modifier flow. Each is checked against targets, and the logic boxes have membership functions to give fuzzy or non-discrete values to the results. The feedback part looks at the most recent -45m fraction ex the circuit and checks if it is within limits. The limits are set by the Local Engineer, but are also affected by the calculated degree of agglomeration. The checks are again fuzzy, and the result is an increase or decrease in the solids gpl target for the agglomerator. This is combined with the feedforward result and after limit checks and a correction for changes in seed variability the new setpoint is downloaded to the DCS.

The benefits are better stability in the precipitation circuit and avoidance of production loss caused by poor control of the agglomeration circuit.

4.10 Mud Filter Wash Optimization

The mud slurry from the wash circuit is filtered on rotary drum filters and re-slurried in condensate before being discharged to the mud stacking area. The purpose is to reduce soda, which improves process efficiency and reduces environmental impact. A key element in achieving this is the displacement wash of condensate applied onto the cake, which then forms part of the filtrate that is returned to the wash circuit. Controlling the filter wash has always been difficult. If the wash is maximized, the mud circuit net wash requirement at some point dictates that the filter wash must be reduced. Having been reduced, it normally is some time before it is again maximized. Normal control schemes are unable to control this situation.

In 1998 AAL developed a scheme on G2 to manage the filter wash. This monitors the net wash, the liquor flow to the mud filters from the last washer, the wash tank level, and the situation in the main process pond. From this it estimates what wash can be added to the filters and distributes this between the wash bars for each of the on-line filters.

The benefits have been seen in reduced soda in the liquor associated with the mud on each filter. This control could not have been achieved by the CRO on his own, nor would it have feasible to implement this type of control scheme on the DCS.


Following the installation of the DCS and other advanced control and information systems AAL has sought to maintain its position at the leading edge of these technologies. This has involved numerous upgrades of software and hardware on the DCS and IT platforms. These upgrades have improved AAL's ability to control and view the process and provided tools for data analysis and reporting. Similar upgrades will continue in the future.

AAL plans to install new applications to improve process operation, efficiency, and quality. Some of these have already been mentioned, notably effluent management. Applications currently under development include supervisory control of the precipitation circuit using model control, boiler load optimization, presentation of cost and efficiency data on-line to all personnel, and development of the Intranet. AAL will also work towards integration of existing systems. Examples are interfacing directly to systems such as Bently Nevada condition monitoring and on-line corrosion monitoring, and allowing G2 access to the maintenance system.

On a more general level AAL's future plans include:

5.1 Transfer all PI Users to PI Process Book

This will provide a single window from which all users can view the process, access logs/reports, and get maintenance information

5.2 Investigate Using Sophisticated Applications to Transfer Information

The most likely mechanism is Object Linking and Embedding (OLE), which allows objects to be used in different applications. OLE for Process Control (OPC) is a standard that has been developed by the OPC Foundation Task Force to facilitate the transfer of information. AAL will assess its suitability for the systems on-site.

5.3 Evaluate Process Modelling

The objective is to use first principle models to show how the process is operating compared to optimal performance. In the past such models have been cumbersome to use off-line and have not been applied on-line. Recent developments offer the opportunity to use on-line modelling. AAL will examine if such products can be used in an on-line application and also how they can interface with G2.