A Self-Adaptive Vibration Reduction Method Based on Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning Algorithm

Round 1

Reviewer 1 Report

In this paper, the dynamic control mechanism of feed-back and feed-forward for Active Vibration Reduction System (AVRS) has been investigated and the Deep Deterministic Policy Gradient (DDPG) algorithm is used to perform real-time self-learning on the system parameters. Adam optimizer is used in the online neural networks of DDPG.

It is suggested that the amplitude curve of transmissibility is appropriate to be set as the reference of reward function since it reflects the modal information about AVRS and is not affected by the form of environmental excitation signal under the same parameters.

For the chosen system simulation and experimental results of transmissibility curve coincide.

Also, simulations and experiments show that the DDPG algorithm can continuously optimize system parameters through interaction with the isolator units to achieve effective suppression of vibrations but with reasonable value selection and optimal calculation of frequency interval to minimise calculation errors.

The use of transmissibility curve for reference of reward function is fundamental and logic approach for vibration problems so together with good experimental results presented in the paper this approach has great potential value in future practical applications.

The manuscript could be considered for publication as it is.

Author Response

Cover letter
Title: A Self-Adaptive Vibration Reduction Method Based on Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning Algorithm
Authors: Xin Jin, Hongbao Ma, Jian Tang, Yihua Kang *
Manuscript ID: applsci-1930063
Dear Editor and Reviewers,
Thank you very much for your valuable comments and suggestions on our manuscript. We appreciate the constructive comments and suggestions on our manuscript. Following the reviewers' comments, we have modified and improved our manuscript according to your kind advice and the referee’s detailed suggestions. 
All co-authors have read and agreed with the revised manuscript. We appreciate for the editors and reviewers’ warm work earnestly, and sincerely hope this manuscript will be acceptable to be published on Applied Sciences.
Thank you very much for all your help and looking forward to hearing from you soon.
Yours sincerely,
Xin Jin
Corresponding author:
Name: Yihua Kang
E-mail: [email protected]

 

Author Response File: Author Response.pdf

Reviewer 2 Report

I would like to congratulate authors for doing such an extensive simulation and experimental research work. DDPG algorithm proved to be an effective method for Active Vibration Reduction System. This method can be considered as an effective tool for other applications also.

Comments for author File: Comments.pdf

Author Response

Cover letter
Title: A Self-Adaptive Vibration Reduction Method Based on Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning Algorithm
Authors: Xin Jin, Hongbao Ma, Jian Tang, Yihua Kang *
Manuscript ID: applsci-1930063
Dear Editor and Reviewers,
Thank you very much for your valuable comments and suggestions on our manuscript. We appreciate the constructive comments and suggestions on our manuscript. Following the reviewers' comments, we have modified and improved our manuscript according to your kind advice and the referee’s detailed suggestions. 
All co-authors have read and agreed with the revised manuscript. We appreciate for the editors and reviewers’ warm work earnestly, and sincerely hope this manuscript will be acceptable to be published on Applied Sciences.
Thank you very much for all your help and looking forward to hearing from you soon.
Yours sincerely,
Xin Jin
Corresponding author:
Name: Yihua Kang
E-mail: [email protected]

 

Author Response File: Author Response.pdf

Reviewer 3 Report

In the paper "A Self-Adaptive Vibration Reduction Method Based on Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning Algorithm" the authors present an interesting method of self-adaptation.

 

The problem approached is well known in literature being an example of school. Including self-adaptive methods based on the natural response of the system are well documented in both academic literature and industrial applications. It is unlikely that at this stage this information will have a technological impact. Every time AI is used to solve once again classic problems successfully solved, I have a feeling of another work with AI.

AI applications would be desirable in cases where classic approaches cannot solve the problem.

The paper is well written, the experimental setup and the presented experiments are well done in accordance with the rules in the field.

 

It is useful to improve the graphic elements in situations where they are practically useless due to the low resolution. Figure 7 is likely to be ordered vertically, and others can be properly enlarged to be sharper.

In my opinion the paper can be published after a minor review.

Author Response

Cover letter
Title: A Self-Adaptive Vibration Reduction Method Based on Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning Algorithm
Authors: Xin Jin, Hongbao Ma, Jian Tang, Yihua Kang *
Manuscript ID: applsci-1930063
Dear Editor and Reviewers,
Thank you very much for your valuable comments and suggestions on our manuscript. We appreciate the constructive comments and suggestions on our manuscript. Following the reviewers' comments, we have modified and improved our manuscript according to your kind advice and the referee’s detailed suggestions. 
All co-authors have read and agreed with the revised manuscript. We appreciate for the editors and reviewers’ warm work earnestly, and sincerely hope this manuscript will be acceptable to be published on Applied Sciences.
Thank you very much for all your help and looking forward to hearing from you soon.
Yours sincerely,
Xin Jin
Corresponding author:
Name: Yihua Kang
E-mail: [email protected]

 

Author Response File: Author Response.pdf