Process simulation - Esko

Question 1

In what ways are the data used?

Answer 1

They are used in modeling, optimization, monitoring, control, maintenance, economic evaluations

Question 2

What kind of errors that can be found in data?

Answer 2

1- Random errors

2- Gross errors

Question 3

What should be done to the data before applying them?

Answer 3

1- data have to be consistent: mass & energy balances

2- possible gross errors have to be detected and accordingly treated

Question 4

What are the steps taken in the process collection and what are its applications?

Answer 4

1- data acquisition
2- data retrieval 
3- data validation and reconstruction
4- data filtering
5- data reconciliation
Applications are:  Parameter estimation - Simulation - Optimization - Advanced control - Accounting - Instrument maintenance

Question 5

Describe random errors?

Answer 5

Normal distribution
▪ describes process data: several
independent phenomena, cumulative
sum of effects
▪ 𝑦𝑚 = 𝑦 + 𝜀
▪ 𝑦𝑚, measured, 𝑦, true value, 𝜀, error, 𝜎2,
variance, 𝜎, standard error
▪ error mean value is zero: 𝐸 𝜀 = 0

Question 6

Describe gross errors?

Answer 6

▪ Non-random events
▪ Causes: miscalibration, malfunction, sensor or equipment fouling
▪ Error has a certain magnitude, the magnitude can
change during time
It has four forms:
1- bias 
2- total failure  
3- drifting
4- precision degradation

Question 7

What are the consequences of errors?

Answer 7

1) The untreated errors can lead to bad plant performance
2) bad control system behavior
3) benefits by optimization are not reached
4) process drifting to an uneconomic or unsafe state
5) results of parameter estimation are wrong

Question 8

What is the purpose of data reconciliation?
what are the methods used in data reconciliation?
and what does it require?

Answer 8

Purpose: To minimize the effects of random errors in the process data before they are used in applications.
Methods:
use process model
− use constraints
− estimate process variables in order to satisfy:
• mass balances
• energy balances
• constraints
It requires redundancy: variables measured and calculated

Question 9

How are the variables classified?

Answer 9

They are classified into 5 types:
1- measured
2- redundant
3- unmeasured
4- observed
5- unobserved

Question 10

What are the requirements of the gross error detection strategies?

Answer 10

1. Detection problem:
   • Ability to detect the presence of one or more gross errors in the data
2. Multiple gross error identification problem:
   • Ability to locate and identify multiple gross errors which may be present simultaneously in the data
3. Estimation problem:
   • Ability to estimate the magnitude of the gross errors

Question 11

What are the gross errors detection methods?

Answer 11

1- Detect outliers. Check adjustments after reconciliation and compare with 95 % confidential region

2. Global test: calculate test statistic and compare with chi2 distribution
3. Nodal test: calculate statistics for each constraint
4. Measurement test: calculate statistics based on adjustments
5. Generalized likelihood test: calculate statistics using the maximum likelihood principle

Question 12

How outliers are detected in gross errors?

Answer 12

1. Perform data reconciliation
2. For each adjustment 𝑎𝑖: if 𝑎𝑖 > 2𝜎 then this measurement is a probable candidate of erroneous measurement
3. Remove bad measurement and repeat the whole process
   ▪ Normal distribution: 2𝜎 is the width of confidence interval where 95% of all measurements should fall.

Question 13

How is a global test is performed?

Answer 13

1. Calculate constraint residual 𝐫 = 𝐀𝐲
2. Calculate variance-covariance matrix 𝐕 = 𝐀𝐖AT
3. Calculate global test statistics 𝛾 = 𝐫𝐓𝐕−𝟏𝐫
4. Compare with 𝑐ℎ𝑖2 distribution with 0.05 = 1-0.95 significance and with degree-of-freedom = independent rows in A. if 𝛾 is bigger, then probably
   there are gross errors
   ▪ Assumption: if there are no gross-errors, then 𝐫 is distributed normally

Question 14

What are the types of process problems? and what are the proposed solutions?

Answer 14

Problems are:
- Simulation problems
- Design problems
- Optimization problems
- Data reconciliation
Solution strategies are:
- sequential modular
- sequential modular with non-linear solution for the tear variables
- equation oriented

Question 15

Describe the sequential modular solver (SM) in Aspen. What kind of problems it has?

Answer 15

It produces solution by performing block by block calculations
- every block solved independently
- when there are recycles, perform iterations until iteration streams (tear streams) do not any more change
 problems
− performs well only in pure simulation problems
− design problems: always an additional iteration loop for each design constraint
− optimization/reconciliation requires special arrangements
− not very flexible

Question 16

Describe the equation oriented (EO) in Aspen. What advantages and disadvantages it has?

Answer 16

In this one, the problem is solved as a large non-linear equation system
 Advantages:
− very flexible problem set-up
− suits well for every problem type
− especially well in problems with heat integration where there are interactions between units
 Disadvantages:
− requires a reasonable initial guess for the solution
− requires advanced numerics