In-the-Loop Simulation Experiment of Aero-Engine Fault-Tolerant Control Technology

Round 1

Reviewer 1 Report

The main topics considered in the paper are:
-    The adaptive turbofan engine model – based on the component-level modelling and extended Kalman filter;
-    Analysis of the fault tolerant control based on analytical redundancy;
-    Analysis of the fault -tolerant control based on switching control rate;
-    Simulation experiments on the sensor fault diagnosis and the sensor fault tolerant control;
-    Hardware in the loop simulation.
The research conducted indicates that good results can be obtained in the sensor fault diagnosis and the sensor fault tolerant control  domain using the approach presented in the paper.
In Section 4 Discussion, the authors have signaled the weak points of the article. The reviewer agrees with the authors' opinion. 
Minor issues:
Line 141:  According to the data provided by NASA  [reference  ?]   ;
Figure 3, Figure 7, Figure 8:  graphs should be presented in uniform scales;
Table 7:  Inconsistent values and units of measure for the variable Error.

Author Response

Thank you for your advice.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper has  grammatical errors and needs minor revisions-

Author Response

Thank you for your advice.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Notations to the paper:

1) Introduction to the paper presents an analysis of the problem and publications in its solution. Otherwise, the tasks of the paper are not formulated. After reading the Introduction, the tasks of the paper are still not understood.

2) The authors do not get the data in Table 1. Thus, a reference to a source of this information is required.

3) The origin of Tab. 2 is also not understood. If they borrowed this information, a reference is needed. If the authors obtain it, they must describe how they did it. The table shows that all four parameters are changed. Then which parameter of the engine, which characterizes the engine power rating, stayed invariable?

4) In Chapter 2.2 (strings 158 and following), the authors give the list of parameters, which they recommend using as the measure ones. However, the paper does not contain information about the considered type of the engine. Therefore, the reader cannot associate these parameters with the engine arrangement.

5) It is not understanding the difference between “total outlet pressure of fan” and “total pressure at inlet of external culvert” (strings 161, 162). Maybe, the engine diagram with a disposition of sensors will help to understand this.

6) In Table 3, the authors give the values of random errors of measurement of parameters, which are used further: error of rotation speed δN=0.15 %, error of pressure δP=0.2 %, and error of temperature δT=0.15 %. As a source of information, the paper of Ogaji S., Li Y.G., Sampa S, et al. [36] is named.

This initiates the following notations:

• Many other works exist, which contain information about errors of the engine parameters measurements, which is determined by real recorded data processing, but the cited work does not belong to them.
• The following values are used in the paper [36]: δN=0.02…0.03 %, δP=0.1 %, δT=0.4 %/ They do not correspond to the values, which are used by the authors of the reviewed paper. Therefore, the reference is not correct.
• The values, which are used by the authors, do not correspond to actual conditions. There is known that the noise of rotation speed, which is measured by a non-contact inductive sensor, is essentially less than the noise of temperature or pressure.

7) Sense and content of Table 5 are not understood:

• The noise is a random value; it is clear that a multiplicity of measurements will contain the multiplicity of the noise realizations, i.e. different values of deviations. Why only one of this random multiplicity is given in the Table?
• Temperature and pressure are not measured in real conditions with precision, which corresponds to the number of signs used for these parameters in Tabl. 5.

8) From string 213 follows that further, the Jacobian J (i.e. a single matrix) is presented. However, in the following string, two matrices are presented: (9) and (10).

9) There is no understanding, why the authors use two equivalent forms of the equations of the dynamic system: (1) and (11) if it is enough to use one of them?

10) Where are the control parameters (fuel flow and nozzle area) in Eq. (1)?

11) The authors use one and the same symbol x for designation of the state vector of the initial system in equations (11), (12), (13), and the extended state vector of the augmented model (14), (15). It will be desirable to use different designations.

12) Description of the example in Chapter 3.1.1 does not contain information about how was the change in the engine technical state simulated? This complicates the understanding of the following examples.

13) Changing of the technical state (strings 271-275) was introduced at 10-th second. In this case, one of the rotation speeds (which is not controlled) must be changed. Most likely, it is Nl. However, this variation at 10-th second is not visible in Fig. 3.

14) Designations in Fig. 3 contain the term “performance degradation”, but really the sensor fault is considered. It is necessary to make corrections.

15) Chapter 3.1.1 does not contain a direct description of sensor faults simulation: for what value the signal was changed, was it constant or variable in time? This complicates the understanding of some elements of the diagrams in Fig. 3. So, for example:

• there is no understanding a sharp decreasing of Nl and Nh at 25-th second in Fig. 3, a and 3, b;
• why values Nl and Nh are increased from 25 to 40 s in Fig. 3, c and 3, d;
• why the dotted line differs from the read line namely at the last segment, if before this segment these lines were coincident?

16) The got values of the gain matrix K of Kalman filter (string 250) show that the last state parameter is unobservable and others are ill-observable because all elements of matrix K in the corresponding strings are tiny. The authors do not analyze this and its influence on the results.

17) Probably, reference to Fig 4 in string 295 is wrong and must be replaced with Fig. 3.

18) Probably, in string 303, “low pressure speed” must be replaced with “high pressure speed”.

19) Fig. 5, a, b, c, d show that the test input at the design and off-design modes were applied at distinct moments of time. However, these actions are described only for the design mode. This complicates the understanding of Fig. 5, a and 5, b.

20) In Chapter 3.1.2, the EPR measuring system is analyzed. However, before (in Chapter 2.3) this parameter was not included in a composition of simulated parameters of the engine. Therefore, it is not understanding how the fault of the measuring system of this parameter is detected if this parameter is not estimated.

21) The paper does not contain information about the method and the results, on which Table 6 is grounded (this table contains the recommendations for selection of switching control rate).

22) Fig. 8, a, b shows that the estimates of rotation speed, which are determined using the Kalman filter, contain a systematic error. This fact is not commented on; the reason and values of the error are not researched.

23) The main result of the paper is a demonstration of the possibility to ensure a fault-tolerant control of the engine using a Kalman filter for state parameters estimation. These results are not original, but they successfully supplement the previously published results of other authors.

Fortunately for other researchers, significant problems are only demonstrated but not seriously studied: precision, stability, and dynamics based on the proposed approach to the fault-tolerant control.

Author Response

Thank you for your advice.Please see the attachment.

Author Response File: Author Response.docx