Control Optimization of a lead Sintering Process

Control Optimization of a lead Sintering Process

16th IFAC Symposium on Automation in Mining, Mineral and Metal Processing August 25-28, 2013. San Diego, California, USA Control Optimization of a le...

4MB Sizes 2 Downloads 26 Views

16th IFAC Symposium on Automation in Mining, Mineral and Metal Processing August 25-28, 2013. San Diego, California, USA

Control Optimization of a lead Sintering Process Marcos Francisco Moreno Molina* and Jorge Morales Castelán* 

*Met-Mex Peñoles, S.A. de C.V., Process Control Department, Torreon, Coah. 27370 Mexico (e-mail: [email protected], [email protected] ). Abstract: An Automation Master Plan was conducted to achieve the control optimization of a blast sintering process based on a Dwight-Lloyd machine. In the pursue of this goal was necessary to use an engineering and scientific approach in order to identify and analyze all the factors influencing the production rate, product quality and environment protection, among others. An assessment was conducted in order to select and implement the appropriate technology, to accomplish the optimization goal. Automation Master Plan has been applied with a scope of five years, and the results achieved, have clearly demonstrated that benefits obtained significantly improved production economics and maximize long-term return on investment.

Keywords: Automation master plan, MPC, expert systems, fuzzy, neural nets, optimization, lead sintering process. 

process variables. The starting point is the process of mixture formation; because lead mixtures determine mostly the way the sintering machine has to be operated to produce sinter as required by the Blast Furnaces process.

1. INTRODUCTION The main goal of the sintering process is desulfurization of the lead ore concentrates to produce a solid and porous agglomerate called Sinter, with uniform size and low-sulfur content. Good quality sinter produced is sent to Blast Furnaces to obtain lead bullion.

Based on the fact that, sintering is a complex operation, combined with our goal of increase productivity, reduce production costs and operate under compliance of environmental government regulations; Process Control department contribution to achieve these goals, was to conduct an Automation Master Plan (AMP) to develop process and technological solutions to reach Sinter Plant aims.

The most important raw material is lead ore concentrates known as galena (lead sulfide). This concentrates are blended with lime, silica, iron, bag house dust and cleanup materials to form the led-silver mixtures. The oxidation process is carry out in a Dwight-Lloyd sintering machine with an area of 120 m2. Depending on the characteristics of the ore used as a charge, the sulfur content is in the range of 12%–15% and lead concentration can be 25%–45%. Sulfur in the mixture provide the fuel for the exothermic reaction and rapidly propagated by a current of air. In this process, sulfides are largely converted to sulfur dioxide (SO2), and the “strong gas” produced from roasting concentrates inside sinter machine is then, after cleaning, further oxidized to sulfur trioxide (SO3), which is converted in strong sulfuric acid as a by-product for sale. The SO2 stream “weak gas”, after cleaning, is converted into ammonium sulfate and sold as a fertilizer.

This paper presents the AMP components, key elements for its development and implementation, as well as, the results and experiences obtained during execution of this project. 2. SINTERING PROCESS AND WHY PROCESS CONTROL AND OPTIMIZATION. The lead ore concentrates are proportioned, blended, granulated, and pelletized into balls called pellets, with the appropriate water content. A shuttle distributor sends the pellets to the ignition and main-bed hoppers, which are fed onto a strand to a depth of 3 cm to form a bed, called the ignition layer and are ignited by using natural gas burners. The ignition temperature is controlled by means of the gas flux. Once the ignition layer is burning, more pellets are fed on top to form a bed of 30-40 cm thick. The strand moves the material along the sinter machine. After four phases (evaporation, heating, reaction, sintering), the pellets become sintering agglomerate with a certain structure, and are discharged from the end of the sintering machine. Goodquality agglomerate is sent to the blast furnaces process while the rest goes through a two-level fragmentation and cooling process and is sent back to the beginning of the process as returned sinter (Tang et al. [1992]).

Because oxidation reaction of lead sulfide is highly exothermic, in order to avoid fusion of the raw material, inside of sinter machine, some amount of produced sinter is crushed, screened and returned to the blending section of the process. Returned sinter along with lead mixture are fed into a pelletizer drum, before it is sent to the machine, in which returned sinter acts as a core in the pellet structure and lead ore adheres to it as the external pellet layer. The continuous process of sintering is a real challenge, because it is a highly interactive, multivariable and sensitive process, where there is a close relationship between all 978-3-902823-42-7/2013 © IFAC

Ideally, a lead sinter machine should be able to produce strong, uniform, low-sulfur sinter at high production rates. 263

10.3182/20130825-4-US-2038.00085

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

The "strong gas" for sulfuric acid production should have a high and uniform SO2 concentration, and the dust content of the sinter should be low. However, a number of difficulties occur in actual operation to prevent these ideal conditions (Knoepke et al. [1982]).

AMP integrated the concept of Digital plant, which implies the evolution of the criteria traditionally considered for the different elements of automation and control, based on the following premises:  Scalable technology, expandable and with integral lifecycle assessment (minimize TCO).

These difficulties can be controlled in a systematic and consistent way through the application of industrial best practices, metallurgical knowledge, measurement, control, and process optimization technologies, to operate the plant according to required production standards.

 Reliable systems, concept).

highly

available

(redundancy

 Interoperability and openness.  Easy to use and maintain (minimize TCO).

Once the process is under control, our next objective is to establish the optimal operating parameters considering the physical, chemical, metallurgical properties of raw material and all the process restrictions including: product quality, production budget, plant capacity, safety and environmental regulations; so the Sinter plant runs as efficient as possible, in order to maximize the amount of mixtures processed.

 Integral set of control engineering tools.  Embedded technology to integrate digital protocol instruments.  Intelligent field devices (smart devices), instead of traditional 4-20 mA.  Embedded technology management.

for

smart

devices

asset

 Embedded global historical record system for process variables. Besides technical aspects, it was necessary to establish the appropriate framework for decision making at the executive level, in order to properly align the AMP within the business plan constraints. While approaching this framework, it was clearly recognized that return on investment time frame will vary according with the type of automation expenditure and therefore the entire project ROI will be a function of it.

Fig. 1. Process profit opportunities from advanced control. Taken from Emerson Global User Exchange, proceedings.

In this context, pragmatic application of Sintering process knowledge, best operation practices, along with the use of technics and technologies of advanced control and optimization, in our opinion, are sine qua non elements in order to achieve the aims established in our AMP.

The pyramids diagram (fig. 3) shows the relationship between type of expenditure, economic benefits expected and the estimated ROI time frame; in our opinion, the figures reported are reasonably accurate, and allow us to prepare proper economic estimations to justify AMP required investment and other resources.

3. AUTOMATION MASTER PLAN Automation Master Plan (AMP) is the process by which automation and optimization technologies are evaluated according to business goals, selected, acquired, implemented, and integrated into the sinter plant for its optimization. The approach to develop the master plan was to establish a methodology that would achieve the objective of optimizing the sinter plant, under a systemic and systematic approach; and aligned at the same time with the strategic business plan.

Fig. 3. Relationship between investment and ROI time frame.

The appropriate alignment with business strategic plan was a critical success factor, because it was emphasized that currently in all industry, but particularly in non-ferrous mineral processing, technical innovation, automation and optimization investment, along with environment protection policies are changing the way companies run the business and make them profitable over the long term.

Fig. 2. Automation Master Plan methodology applied.

264

IFAC MMM 2013 August 25-28, 2013. San Diego, USA



In this context, AMP along with digital plant concept, leveraged by advanced control and optimization technologies, incorporate intelligence to every stage of production process, leading to improvements in terms of sintering product properties, production rate, cost, maintenance, energy and materials consumption, operating accuracy, waste and environment protection management.

      

3.1 Execution of Automation Master Plan (AMP). Proper execution of AMP starts with an interdisciplinary team approach. The organization should carefully consider the structure of the interdisciplinary team, which includes key staff members from the specialties of plant operations, process engineering, plant maintenance, plant engineering, scientific research center, six sigma, instrumentation, process control engineering, and other staff departments.



Ore lead concentrates proportioning and blending system. Charge mixing and conditioning system. Burn-trough point based on gas temperature profile inside machine hood. Return sinter crushing and screening system. Return sinter conditioning system. Ventilation system. Automatic plant start-up sequence. Diagnostic of abnormal operating conditions (Expert system with Fuzzy Logic and Neural Networks). Optimization model for sintering machine.

All other important outcomes from AMP are as follows: 

All team members should utilize their specialized knowledge and skills to achieve the main goal of optimize sintering process control. The following factors have to be considered by team members:

 

 Development of the automation master plan considering, the operational conditions, constraints (environment protection), and its cost.  Implementation of the automation master plan providing all the necessary components during the project execution phase.  Make sure that all process control and optimization findings have been properly implemented during all phases of design, and during the implementation.  Cooperate in the continuous improvement of AMP development tasks.

    

Estimated economic and operating benefits from process control and optimization improvements. Required project investment, including capital expenditures, cash flow and ROI. Definition of all phases required for execute AMP strategy. AMP approval from executive management. Integrate execution team members. Develop the project implementation schedule. Process control system design and specifications. Selection of process control and optimization technology. 4. SINTER PLANT SIX SIGMA RESEARCH.

The six sigma research was conducted to determine in a precise form, the pragmatic cause-and-effect relationship existing among the different sintering process variables, classify them, quantify them and determine the impact that directly or indirectly have over the treatment of mixtures, sinter quality, quantity produced and environment protection.

 Prepare progress and cost reports. Master planning process should be conducted in a manner which maximizes long-term investment effectiveness and project implementation. For that reasons, AMP should be concise and designed for easy understanding, implementation, and execution. Successful implementation of the AMP starts with a well-organized and realistic plan.

Goal statement in the six-sigma team charter was as follows: “Identify and analyze the required operating conditions to optimize sintering process, given any mixture for treatment and subject to the applicable constraints (environment protection, plant capacity, process capability, mechanics, quality, quantity, opportunity, cost, etc.), reducing process variability, as well as, achieving constant production rate and improving sinter quality.” AMP aligned with business goal: Keeping in mind, that our business goal, while optimizing sinter plant, is producing good quality sinter to be processed at blast furnaces process, in the most efficient way and at the lowest possible cost; in order to produce lead bullion under the required quality and quantity, so as to produce from it, the final refined valuable products, which are in the end, sent for sale to the metals market (silver, gold, lead, and others).

Fig. 4. Interdisciplinary matrix for AMP execution.

The main outcome from the development of AMP was the identification of key elements for sinter plant control and optimization, among which are the following: 265

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

composition within specific ranges while complying with the calorific power required for sintering the mixture.

4.1 Six sigma research obtained results. The six sigma research was conducted based on the very well-known DMAIC methodology (Gygi et al. [2005]):

In the mixture is very important to maintain a constant and appropriate amount of calorific power, which is a function of the quantity and type of sulfides. Operation of sinter machine requires also a reliable control of the proportioning system, for the proper amounts of mixture and returned sinter being dispensed in order to comply with sulfur content that can be eliminated in one pass over the sinter machine; we call this condition as a sulfur head to be fed into the machine.

 Define: Set the context and objectives for the project.  Measure: Get the baseline performance and capability of the process being improved.  Analyze: Use data and tools to understand the causeand-effect relationship in the process.  Improve: Develop the modifications that lead to a validated improvement in the process.  Control: Establish plans and procedures to ensure the improvements are sustained.

Types of sulfides compounds contained in mixture and their oxidation reactions: 2PbS + 3O2  2PbO +2SO2 + 199.6 kcal

4.1.1 Model of feasible operating point.

3ZnS + 3O2  2ZnO + 2SO2 +212.6 kcal 4FeS2 + 11O2  2Fe2O3 + 8SO2 + 791.7 kcal

What we called, model of feasible operating point (FOP), is a statistical calculation to determine which is the pragmatic operating condition, regarding the amount of metric tons per day of mixture treated in sinter machine, which will be achievable if we reduce process variability by mean of process control technology. Based on this figure, we were able to set a baseline towards our optimization goal.

Other important factor is the returned sinter and pellet sizes, as well as proper distribution pellet size after the blending and conditioning systems. In conjunction these factors have a strong influence in bed permeability and packing inside the machine, which contributed to reaction front through the bed and therefore on the oxidation reaction efficiency.

The calculation is based on the following expression:

𝐹𝑂𝑃 = 𝑋 + (𝑈𝐿 − 𝑋) × (1 − 𝑆𝑎𝑝𝑐 /𝑆𝑡𝑜𝑡 ) Where

Within conditioning system is also very important to maintain a proper amount of moisture (humidity) on the produced pellet before sending them to ignition section. Humidity has an important role during the oxidation phase heat front transfer process.

[1]

𝑋 = Average amount of mixture treated, t/d.

UL = Operating Upper limit achieved, t/d.

Front flame speed (FFS) is limited by the amount of available energy contained in the pellet; it influences the machine speed and the air flow required for reaction, and in consequence the residence time of charge inside the machine.

Stot = Actual standard deviation of amount of mixture treated, t/d. 𝑛 𝑖=1

𝑋𝑖 − 𝑋 𝑛−1

𝑆𝑡𝑜𝑡 =

2

A homogenous oxidation reaction of bed pellets depends on an adequate burning on ignition layer section of the machine.

Sapc = Expected standard deviation of amount of mixture treated, once advanced process control is in place, t/d. n

2

S

apc

 S

cap

2

 S cap   S  S cap  tot  ,





i2

(X

i

 X

i 1

All these factors, and more, have been reported by many other studies conducted for lead-zinc sintering machines. See references 1, 2, 3 and 4.

)2

2 ( n  1)

In our case, FOP resulted in reasonable amount of mixture treated by day basis of 1,290 t/d. Based on this calculation our baseline was established, and all the technical analysis were conducted focused on it. Therefore the process control and optimization solutions should be aligned with it. 4.1.2 Sintering process variables relationships. From the operations perspective, the main goal is maximize mixture oxidation as much as possible along with maximize the amount of mixture treated but within the quality, quantity and environment protection constraints. The mixture density varies as a function of the different components, which in our case, is an important source of variability; for this reason we have to maintain the mixture

Fig. 5. Sintering variables relationship diagram.

266

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

Figure 5 shows a diagram representing all the sintering process variables relationship, which is a graphical summary of the six sigma research findings. 4.1.3 Sintering process optimization. Model balance of sintering machine is defined as follows: mt = (hc×am)Vm×da×%m

[2]

Where: mt= tons of mixture treated per time unit, t/h.

Fig. 6. Flame Front Speed function.

Vm = sintering machine speed, m/min. hc = main bed height, m. am = sintering machine width, m. da = bed pellets bulk density, kg/m3. %m = percentage of mixture on pellet. From equation [2], we concluded that for maximizing the amount of treated mixture, it is required to maximize the speed machine and %m. Bed pellet bulk density is determined by pellet elaboration process and its properties, like consistency, which modifies the contact points between pellets and becomes a crucial factor for the front flame speed (FFS).

Fig. 7. Apparent optimum speed machine.

Because the machine length is fixed, speed machine (Vm) has to be defined by the FFS (combination of heat transfer front and reaction front). Based on this, we defined the concept of “Optimum machine speed” as the speed in which ignition process is completed within the length of the machine; therefore allowing maximum utilization of the machine length.

In complement, we analyzed machine operating conditions under several qualities of actual pellets, and it was also observed that exists an optimum apparent point for machine speed which closer to the former estimate of 1.15 m/min; in this case we obtained a value of 1.13 ± 0.03 m/min. Under this analysis we observed that the apparent optimum is located within a window frame at which the different qualities of pellets group themselves. These facts are presented on figure 8.

One way of determining approximately FFS is, identifying the ignition point (wind box) which corresponds at maximum temperature (inside the machine) taken from the temperature gas profile in the machine hood. We can then write-down the following expression: FFS ≈ hc / (LTmax/Vm)

[3]

Where: LTmax = machine wind box position with highest temperature. In practice, the plant operator does not know FFS in advanced; therefore the operation optimization procedure is based on finding the synchronization point between Vm and FFS for a specific mixture. For that reason the optimization variable is the machine speed based on this factor (see fig. 6). Based on all factors mentioned before, we were able to determine de apparent optimum speed, this is graphically represented in figure 7, where we used actual plant data to analyze under which conditions and it is possible to approach a maximum point if mixture treated and the corresponding speed machine.

Fig. 8. Apparent optimum machine speed with different pellets.

Finally, including all cases of operating conditions selected, we graphically modeled the “optimization zones”, including the sulfur content on the pellet charge fed into the machine, please refer to figure 9.

267

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

Process control and optimization strategy was implemented considering a logical sequence, in which basic regulatory control and advanced classical control were first developed. After a maturity period of understanding and stabilization, advanced control took place applying mainly model predictive control, just in some of critical points of control, where higher precision and stability were required.

29.20

46

31.22

44

33.24

42

35.26

40

37.28

CC

38

39.29

36 34

41.31

32 30

43.33 45.35

28 26 24 22

Optimization technology was implemented at the top of the control system, in order to send the optimized operating setpoints according to the actual operating constraints.

7 .5

20 1 .0

V1

7 .0

.1

1 .2

1.3

S

6 .5

Fig. 9. Optimum zones for machine speed, mixture treated and head sulfur content on pellet.

Based on previous analysis, on the fact that process response is asymmetric, and taking into account that pragmatic optimization has to consider the point at which the sinter product complies with quality specifications (see fig 5), we proposed the following operating conditions for machine speed: Table 1. Machine speed proposed limits. Limits

Fig. 12. Control and Optimization technology layers.

Vm, m/min

Upper specification limit (USL) Upper control limit (UCL) Apparent Optimum

(AO) Lower control limit (LCL) Lower specification limit (LSL)

5.1 Model Predictive Control (MPC) applied on Sinter plant.

1.21 1.16 1.13 1.1 1.08

Sinter plant, as we have discussed previously, has at least five key control points: a)

Proportioning and charge demand.

b) Pellet moisture.

5. SINTER PROCESS CONTROL.

c)

The process control and optimization strategy is based on the operational excellence strategy, as shown in the following diagram:

Ignition bed temperature and suction fan flow.

d) Return sinter cooling and binder addition. e)

Air flow for sulfides oxidation.

In order to achieve a more stable, robust and precise control at those key points, we decided to use an MPC strategy for each one. Fig. 10. Operational excellence strategy.

The reasons to select MPC as a control technology were:

Concept developed by Emerson Process Management ©

a)

According to digital plant concept core elements, the technology decision was made, and the process control architecture, measurement technology, communication protocols and optimization platform were implemented according to the following scheme:

Ability to handle multivariable systems.

b) Ease of handling complicated dynamics (e.g., high dead times, large time constants, etc.) c)

Ability to manage constraint handling with a systematic approach; pushing the plant towards its limits.

d) Plain applicability of feed-forward (measurable disturbances). e)

Predict future states and track reference trajectory.

f)

The DCS technology selected has embedded MPC, which actually runs on the controllers, instead of running on a separate personal computer.

The key success factor was to stratify the MPC application, instead of trying to implement a “big” MPC strategy; we realized that it would be a better approach to have individual MPC units that would be easier to design, understand, train for and maintain to keep them running with high availability. This strategy proved to be useful and successful.

Fig. 11. Control and Optimization technology architecture. 268

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

Another important consideration was the approach to reduce the dependence of MPC experts or consultants. This is important, to achieve the aim to keep MPC running with high availability, so the plant process control engineers have all the necessary knowledge and skills to perform this main task. Our approach is illustrated in the following scheme:

Fig. 15. DCS MPC function block diagram. Emerson Process Management, DeltaV books on-line.

6. SINTER PLANT CONTROL OPTIMIZATION. Fig. 13. Strategy for APC ongoing support and high availability. Modified from Process Control: A practical approach by Myke king. Page 388. Wiley, 2011.

Control and optimization are terms that sometimes are erroneously interchanged. Process control has to do with adjusting manipulated variables to keep process controlled ones, close to their specified set-points or reference points.

Implementing MPC on Sinter plant was a challenging task, because all the main flow materials are solids and therefore we have no control valves, but conveyor belts and weighting systems. The best case was using variable frequency drives as final control elements for controlling motors.

However, process optimization, through the application of specialized mathematical techniques, helps us to determine those set-point values, such that the process operates as closer as possible to its “best” profitable conditions, within the restrictions that imprints process constraints that surrounds industrial processes.

With the use of MPC a new feature was introduced at operations control room, Process Variables Prediction. Now plant operators have a future reference for the controlled variable, which is not available when using PID controllers. This improvement is useful when the SP is changed since operators got an on-line prediction of the process variables trajectory towards reaching the new SP, or in the case where a process disturbance takes place the PV prediction gives operators the confidence that PV will be controlled and how long will it take for the PV to stabilize again.

Fig. 16. Optimization concept.

6.1 Optimization design and implementation. Main process optimization strategy focus on maximize the amount of treated mixture, within the process constraints of sulfur content in sinter produced and the SO2 emission limits. Solution is based on an optimization algorithm guided by a mathematical process model. A progressive implementation approach is used together with operations staff training program to make the optimization application easy to understand, use and maintain.

Fig. 14. MPC Reference trajectory, Prediction & Control horizon.

MPC technology implemented have proved to be very robust and reliable, easy to use and maintain, and with high availability due to its characteristic of running in the DCS controllers. Following an MPC functional block diagram is shown for reference.

The different stages of implementation are as follows: a) Monitoring of key decision variables. b) Training of the predictive model.

269

IFAC MMM 2013 August 25-28, 2013. San Diego, USA

d) Complete optimization for decision variables.

 Optimum set-points (proportioning system, conditioning, etc.).

e) Assisted optimization off-line.

 Optimum machine speed.

f) Full optimization on line.

 Flexibility to adjust the target set-point of final sulfur content in sinter product, according to economic opportunities at blast furnace process.

c) Pre-optimization stage using advice from sintering process operators and experts.

A dynamic discrete system was used as a process model, with the following characteristics: 1.

Key point process parameters were codified in a vector type X(t), and monitored.

2.

Predicting vector values in time (t+dt):

3.

Components are vector functions representing process materials and physical parameters characteristics on certain points of time.

4.

Each function component must be calculated based on historical process data and individually. Data analysis helps to define the more appropriate family models, which it could be represented as a simple linear model or may be a neural network.

for MPC’s controllers pellet and return sinter

 Push the operating conditions towards its environmental constraints, but without trespassing them.  Flexibility to drive optimization changing priority between production and quality.  Obtain an estimate of economic impact of mixture composition variability.  Accountability for the production level and its limits.

For example, in order to obtain an estimated value for head sulfur content as input to the sinter machine, we can expect it to be an aggregated function of mixture sulfur content and return sinter. Parametric family is expected to be like this: Fig. 18. Actual optimization application operator screen.

The weight vectors (’s, ’s) are calculated by mean of real process data for the prediction to be reliable. The time delays between them, which are inherent to the process, should be taken into account.

7. CONCLUSIONS Automation Master Plan probes to be an efficient and effective methodology to execute the sinter optimization project, along with experts’ process knowledge, six sigma, process control technology and optimization, by mean of which we were able to achieve the operating improvement and optimization of sintering process.

Therefore, according with all those concepts, the objective function to maximize the amount of treated mixtures can be expressed as:

REFERENCES

We can express this concept in the next diagram:

X.R. Tang, D.Y. Wang, and Q.C. Zhang (1992). Sintering theory and technology. Central South University of Technology Press, Changsha. (1) John R. Knoepke, Hung-Yang Tsai, and Arthur E. Morris. Factors influencing the production rate and quality of lead sinter. American Society for Metals and the Metallurgical Society of AIME. Metallurgical Transactions B, volume 13B: 15-29, March, 1982. (2) Chun-Sheng Wang, Min Wu, Jin-Hua She, Wei-Hua Cao and Yong He. Qualitative and quantitative synthetic methodology for blending optimization in lead-zinc sintering. Proceedings of the 17th world congress, IFAC. Seoul, Korea, July 6-11, 2008. (3) Min Wu, Chen-Hua Xu, Yu-xiao Du. Intelligent optimal control for lead-zinc sintering process state. Transactions of Nonferrous Metals Society China, 16 (2006) 975-981. (4) Craig Gygi, Neil DeCarlo, Bruce Williams (2005). Six Sigma. Chapter 3. Wiley Publishing, Inc., Indiana, USA. (5)

Fig. 17. Objective function graphic.

The expected results from the optimization application are:  Prediction of environmental emissions and sinter quality and production. 270