A Hybrid Approach: Dynamic Diagnostic Rules for Sensor Systems in Industry 4.0 Generated by Online Hyperparameter Tuned Random Forest (Version 1, Original)
|Reviewer 1 Carlos Silva Federal University of Paraná (UFPR), Electrical Engineering Department, Curitiba, Brazil||Reviewer 2 Yannis Karnavas Democritus Univ Thrace, Dept Elect &amp; Comp Engn, Elect Machines Lab, GR-67100 Xanthi, Greece||Reviewer 3 Yinsheng Chen Institute for Bioengineering of Catalonia|
Approved with revisions
Approved with revisions
Mallak, A.; Fathi, M. A Hybrid Approach: Dynamic Diagnostic Rules for Sensor Systems in Industry 4.0 Generated by Online Hyperparameter Tuned Random Forest. Sci 2020, 2, 61.
Mallak A, Fathi M. A Hybrid Approach: Dynamic Diagnostic Rules for Sensor Systems in Industry 4.0 Generated by Online Hyperparameter Tuned Random Forest. Sci. 2020; 2(3):61.Chicago/Turabian Style
Mallak, Ahlam; Fathi, Madjid. 2020. "A Hybrid Approach: Dynamic Diagnostic Rules for Sensor Systems in Industry 4.0 Generated by Online Hyperparameter Tuned Random Forest." Sci 2, no. 3: 61.
Article Access Statistics
Federal University of Paraná (UFPR), Electrical Engineering Department, Curitiba, Brazil
In this work, a hybrid component Fault Detection and Diagnosis (FDD) approach for industrial sensor systems is established and analyzed, to provide a hybrid schema that combines the advantages and eliminates the drawbacks of both model-based and data-driven methods of diagnosis.
The paper provides a new contribution where it is possible to create a hybrid FFD for industrial sensors. The contribution is clear and inedita. Some questions are indicated as showed bellow.
- In the first paragraph, industry 4.0 application examples can be supported by a relevant quote from the area where application examples are used.
- A sentence "three main categories; sensor faults, actuator and component/system faults", the authors describe different classes of faults. The sensor and actuator are ok, but about component/system as shown below It is needed to be revised. This type of fault, maybe, must be separated because It is possible having a fault in hardware systems such as a high temperature or low-efficiency processing of CPU. IN the other case, can be a fault due to bad implementation or coding the software.
- In "healthy state" it is not common in this case, could use "normal state".
- The final of section 2 where is presented a brief of related works it is suggested to insert a comparative table where the authors can light the real contribution of the proposed method, in order to show vantages and disadvantages.
- The sentence "In order to understand the mathematics behind RF, it is highly recommended to go through the explanation of decision trees and how they work in the first place." could be deleted.
- Equation of table 1 could be better presented as an equation in the text. What is the source citation of these equations?
- Algorithm 1 doesn't sound being an algorithm, but a description of steps. It is suggested change for algorithm steps or uses another way presentation.
- Section 4 is very short and doesn't bring important information such as (i) why is it used this source data from ? (ii) Is it possible to describe this machine plant test as a block diagram?
- Subsection 5.1 sounds to be part of the Material and Methods. Each block of figure 1 needs to be in deep described, especially where it is necessary to provide the type information used each one (input and output variables).
- At "The following table," the text could be removed.
- Based in Table 3, is it possible to affirm the method based on RF to improve techniques to identify faults, but no more than CART? Can the authors clarify what characteristics RF is better, such as, processing data information?
- Table 5 and also text present information about the accuracy of RF, but the accuracy is less than 1%. Is it correct? These values sounds are very low. 0.985% or 98.5%?
- In Figure 2 must be inserted x and y-label.
- Must insert the conclusion section at the end.
- At the Abstract and other parts of the manuscript, it is informed SQL queries was used. SQL is a language for database (DB) systems where is possible to create and select information from tables from DB, but It was not described the queries used such as "select from *". What kind of DB used?
Response to Reviewer 1Sent on 20 Sep 2020 by Ahlam Mallak, Madjid Fathi
Democritus Univ Thrace, Dept Elect &amp; Comp Engn, Elect Machines Lab, GR-67100 Xanthi, Greece
In this work, the architecture of a hybrid FDD method, between model-based and data-driven approaches to achieve FDD for component faults, is introduced. In this hybrid method, the datadriven part is represented by an optimized and hyperparameter tuned RF, in order to generate dynamic, diagnostic graphs that are later converted into a set of diagnostic rules and fed into a predefined system diagnostic model, acting as the model-based part of the proposed FDD system. The proposed approach provides a dynamic solution, unlike other model-based FDD approaches.
The work is technically sound and the paper is well written.
Response to Reviewer 2Sent on 20 Sep 2020 by Ahlam Mallak, Madjid Fathi
Institute for Bioengineering of Catalonia
This paper presents a hybrid method for sensor array fault diagnosis.
1.Industry 4.0 is an important development direction, which is worth studying. It is suggested to explain some content of Industry 4.0 in the Abstract, because industry 4.0 is mentioned in the Keywords section.
2.Can the author explain the difference between the fault diagnosis of traditional sensor system and the fault diagnosis of sensor system in Industry 4.0? Can the traditional sensor fault diagnosis method be applied to the fault diagnosis of sensor system in Industry 4.0?
3.The proposed method is consistent with the fault diagnosis method based on pattern recognition in essence.Data-driven methods can make data selection, such as PCA. The random forest method is equivalent to a classifier to realize the classification of features.What are the differences between the proposed method and pattern recognition?
4.The resolution of some pictures in this paper is low, and the space is too large, please adjust the author.For example, FIg.3 and FIg. 4.