Available online at www.HighTechJournal.org  

HighTech and Innovation  
Journal 

Vol. 6, No. 3, September, 2025 

 

 

863 

 

ISSN: 2723-9535 

 

FATA-ResNet Network for CAD/CAM Integration in Cloud 

Manufacturing 

Chao Gu 1*  

1 School of Intelligent Manufacturing, Jiaxing Vocational and Technical College, Jiaxing 314036, Zhejiang, China. 

Received 13 May 2025; Revised 23 July 2025; Accepted 07 August 2025; Published 01 September 2025 

Abstract 

This paper focuses on the application of mechanical engineering CAD/CAM integration technology under the cloud 

manufacturing framework, aiming at solving the current technical integration problems in manufacturing informatization. 

The study analyzes the demand and current situation of 3D CAD/CAM integration in a cloud manufacturing environment, 

combines the mirage optimization algorithm (FATA) and residual neural network (ResNet), and proposes a CAD/CAM 

integration application analysis model based on the FATA-ResNet network. Firstly, the functional requirements of 

CAD/CAM technology integration in a cloud manufacturing platform are clarified, including 3D model uploading and 

downloading, process file generation, and cross-platform data sharing. Then, the hyperparameters of the ResNet network 

are optimized by the FATA algorithm to improve the accuracy and efficiency of the model in integration application 

analysis. The experimental results show that the FATA-ResNet model outperforms the traditional model in terms of 

accuracy, recall, and F1 score while possessing faster convergence speed and higher computational efficiency. In addition, 

the operation modules in the cloud platform, including the task management interface and 3D process editing function, 

were designed and validated, further demonstrating the practicality of the method. Future research will focus on the 

validation of multi-scene data, model resource optimization, and real-time collaborative operation to promote the in-depth 

application of CAD/CAM technology in intelligent manufacturing and provide support for the digital and intelligent 

development of manufacturing. 

Keywords: Cloud Manufacturing; Mirage Optimisation Algorithm; Residual Neural Networks; CAD/CAM Integration Techniques. 

 

1. Introduction 

In recent years, the manufacturing industry has undergone significant transformation with the advent of Industry 4.0 

and initiatives like "Made in China 2025" [1]. These advancements have shifted the focus from mere product 

manufacturing to delivering high-value-added products and services [2]. As a core sector within manufacturing, 

mechanical engineering relies heavily on information technologies such as 3D CAD/CAM to shorten product design 

cycles and enhance market responsiveness. 

Cloud manufacturing, which leverages cloud computing to virtualize manufacturing resources and capabilities, has 

emerged as a promising paradigm [3]. It combines technologies like IoT, big data, and AI to improve efficiency, reduce 

costs, and increase flexibility [4]. The integration of CAD/CAM technologies within this framework is crucial for 

achieving collaborative and integrated manufacturing processes. 

                                                           
* Corresponding author: 13957303929@163.com 

 
http://dx.doi.org/10.28991/HIJ-2025-06-03-08 

 This is an open access article under the CC-BY license (https://creativecommons.org/licenses/by/4.0/). 

© Authors retain all copyrights. 

https://creativecommons.org/licenses/by/4.0/
https://orcid.org/0009-0008-8380-1351


HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

864 

 

Numerous studies have explored CAD/CAM integration [5-7]. For instance, researchers have proposed various 

methods such as data exchange interface-based integration, collaborative design-oriented integration, PDM-based 

integration, and manufacturing feature-based integration [8, 9]. However, despite these efforts, several limitations 

persist. First, the integration of soft manufacturing resources often fails to fully utilize the functionalities of various 

software modules. Second, there is inadequate research on the integration and evaluation of 3D CAD/CAM technologies 

specifically within the cloud manufacturing context [10]. 

To address these gaps, this study proposes a novel approach. A CAD/CAM integration analysis model is developed 

by combining the Mirage Optimization Algorithm (FATA) with the Residual Neural Network (ResNet). The Mirage 

Optimization Algorithm (FATA) is employed to optimize the hyperparameters of the ResNet architecture. In this study, 

no modifications are made to the core structure of the standard ResNet. Specifically, ResNet50 is utilized in its original 

form, including its standard layers and connections. The FATA algorithm was only used to optimize ResNet's 

hyperparameters, with no changes to ResNet's internal architecture. The FATA algorithm uses specific parameters and 

symbols to represent different aspects of the optimization process. This model not only enhances the accuracy and 

efficiency of integration analysis but also offers faster convergence and improved computational performance [11]. By 

designing and validating operation modules within a cloud platform, we demonstrate the practical applicability of our 

method [12]. This research aims to advance the application of CAD/CAM technologies in intelligent manufacturing and 

provide robust support for the digital and intelligent transformation of the manufacturing industry. 

This paper constructs a comprehensive system based on the IASB framework and enhanced with a PO-BP model. 

Section 2 focuses on the theoretical foundation and construction of the data asset accounting system, including 

recognition, measurement, recording, and reporting. Section 3 presents the integration of the Political Optimizer (PO) 

algorithm with the BP neural network to develop a data asset valuation model. Section 4 offers a comparative analysis 

using open-source datasets to validate the model’s performance against traditional algorithms. Finally, Section 5 

concludes with a summary of findings, acknowledges the limitations, and proposes directions for future research. This 

structured approach ensures a thorough exploration of both conceptual foundations and practical implementations, 

offering valuable insights into data asset accounting in the digital economy. 

2. CAD/CAM Integration Technology in Cloud Manufacturing Framework 

2.1. Status of Research 

In recent years, the concept of cloud manufacturing has attracted international academic attention, and many countries 

have researched cloud manufacturing [13]. Cloud manufacturing is a new manufacturing model based on cloud 

computing technology, which virtualizes and services manufacturing resources and manufacturing capabilities and 

provides them to users through the Internet [14], as presented in Figure 1. It combines cloud computing, the Internet of 

Things, big data, artificial intelligence, and other technologies, aiming to improve manufacturing efficiency, reduce 

costs, and enhance the flexibility and responsiveness of manufacturing systems. 

 

Figure 1. Cloud manufacturing 

The main features of cloud manufacturing include 1) resource virtualization; 2) servitization; 3) on-demand 

customisation; 4) flexibility and scalability; 5) data-driven; and 6) remote monitoring and maintenance, as shown in 

Figure 2. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

865 

 

 

Figure 2. Cloud Manufacturing Characteristics 

For cloud manufacturing, all domestic and foreign studies have achieved remarkable results. The concept, 

background, architecture, and technical features of cloud manufacturing have been described. Its application in the 

aerospace R&D process has been proposed to reduce informatization costs and enhance efficiency. Platforms 

integrating various CAD/CAM/CNC interfaces have been developed to resolve CAX compatibility issues. A cloud 

manufacturing model based on the STEP standard for process collaboration and data integration has also been 

investigated [17-20]. 

CAD/CAM integration research and development (Figure 3), is the core link of manufacturing information 

technology, but also to achieve an important part of cloud manufacturing, many developed countries have always 

attached great importance to the integration of CAD/CAM. CAD/CAM integration methods are mainly the following 

four, as shown in Figure 4, specifically including: 1) data exchange interface-based integration technology [21]; 2) 

collaborative design-oriented integration technology [22]; 3) PDM-based integration technology [23]; 4) manufacturing 

feature-based integration technology [24]. 

 

Figure 3. CAD/CAM integration technology concept 

 

Figure 4. CAD/CAM integration method 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

866 

 

2.2. Needs analysis 

The cloud manufacturing platform provides users with CAD/CAM and other soft manufacturing resources data and 

information calls are an important part of the cloud manufacturing platform information technology services [25]. 

Achieving the integration service of 3D CAD/CAM and PDM is an important part of product co-design and process co-

design, and the specific integration module requirements are shown in Figure 5, which include the following: 1) CAD 

parts information extraction; 2) uploading and downloading of 3D CAD models, and intelligent loading and generation 

of 3D process files; 3) 3D CAM process information extraction; 4) 3D CAM and CAPP integration; 5) manufacturing 

information browsing on mobile devices [25]. 

 

Figure 5. Requirement analysis of CAD/CAM integration based on cloud manufacturing 

2.3. Architecture Analysis and Design 

The cloud manufacturing platform system is composed of a resource layer, an intermediate layer, a core functional 

layer, a platform portal layer, and a service application layer, as shown in Figure 6. 

 

Figure 6. Cloud manufacturing application architecture 

The mechanical engineering CAD/CAM integration and integration framework for cloud manufacturing is shown in 

Figure 7. From Figure 7, in the cloud manufacturing platform, digital design software such as CAD/CAM and PDM are 

integrated to provide technical support for the management and transfer of data and models in the process of collaborative 

production, and by uploading the data generated by CAD/CAM and so on to the cloud database through PDM, it can 

provide data support for the product's full life cycle design. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

867 

 

 

Figure 7. CAD/CAM integration framework 

2.4. CAD/CAM Technology Integration Application Analysis 

According to the CAD/CAM integration and integration design ideas, this paper takes the design function value as 

input and the integration and integration test value as output to construct the CAD/CAM technology integration 

application analysis model, as shown in Figure 8. To improve the integration technology application analysis efficiency, 

this paper adopts a machine learning algorithm, through learning training, to construct a CAD/CAM technology 

integration application analysis model, and then uses an intelligent optimization algorithm to optimize the model to 

improve. 

 

Figure 8. CAD/CAM Integration Application Analysis Model Input and Output 

3. Mirage Optimization Algorithm 

The Mirage Algorithm (Fata Morgana Algorithm, FATA) [9] is a novel population intelligence optimization 

algorithm proposed in 2024, which is inspired by the mirage formation process in natural phenomena as shown in Figure 

9. The FATA algorithm proposes two core strategies by mimicking the propagation of light in an inhomogeneous 

medium --Mirage Light Filtering Principle (MLF) and Light Propagation Strategy (LPS) to optimize the search process 

and enhance the algorithm's global search capability and local exploitation. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

868 

 

 

Figure 9. Principle of the FATA algorithm 

3.1. Initialization 

As with the other algorithms, random initialization is used: 

𝑥𝑖 = 𝑟𝑎𝑛𝑑 ⋅ (𝑈𝑏 − 𝐿𝑏) + 𝐿𝑏  (1) 

where 𝑈𝑏 denotes the upper bound of the optimization problem, 𝐿𝑏 denotes the lower bound of the optimization problem 

and 𝑟 and denotes the random number. 

3.2. Mirage Filtering Strategy 

In the physical process of mirage formation, objects emit two types of light. Most light rays belong to the first type 

(other rays), which do not propagate and form mirages. The other type of light undergoes a physical change to form a 

mirage and is called mirage light (Figure 10). The specific mathematical model is calculated as follows: 

 

Figure 10. FATA algorithm mirage filtering strategy 

𝑥𝑖
𝑛𝑒𝑥𝑡 = {

𝐿𝑏 + (𝑈𝑏 − 𝐿𝑏) ⋅ 𝑟𝑎𝑛𝑑 𝑟𝑎𝑛𝑑 > 𝑃
𝑥𝑏𝑒𝑠𝑡 + 𝑥𝑖 ⋅ 𝑃𝑎𝑟𝑎1 𝑟𝑎𝑛𝑑 ≤ 𝑃&&𝑟𝑎𝑛𝑑 < 𝑞

𝑥𝑟𝑎𝑛𝑑 + [0.5 ⋅ (𝛼 + 1)(𝑈𝑏 − 𝐿𝑏) − 𝑎𝑥𝑖] ⋅ 𝑃𝑎𝑟𝑎2 𝑟𝑎𝑛𝑑 ≤ 𝑃&&𝑟𝑎𝑛𝑑 ≥ 𝑞
  (2) 

𝑃 =
𝑆−𝑆𝑤𝑜𝑟𝑠𝑡

𝑆𝑏𝑒𝑠𝑡−𝑆𝑤𝑜𝑟𝑠𝑡
  (3) 

𝑞 =
𝑓𝑖𝑡𝑖−𝑓𝑖𝑡𝑤𝑜𝑟𝑠𝑡

𝑓𝑖𝑡𝑏𝑒𝑠𝑡−𝑓𝑖𝑡𝑤𝑜𝑟𝑠𝑡
  (4) 

where 𝑥𝑖  denotes a ray individual, 𝑥𝑖
𝑛𝑒𝑥𝑡  denotes a new ray individual, 𝑃 denotes a ray population quality factor, 𝑞 

denotes an individual quality, 𝑆 denotes a population quality, 𝑆worst  denotes the worst population quality, 𝑆𝑏𝑒𝑠𝑡 denotes 

the best population quality, 𝑓𝑖𝑡𝑖 denotes the ith ray fitness value, fit 𝑡best  denotes the optimal individual ray fitness value, 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

869 

 

𝑓𝑖𝑡worst denotes the worst individual ray fitness value. The parameter 𝑃𝑎𝑟𝑎1 represents the first stage refractive index, 

which is initially set to 0. The parameter 𝑃𝑎𝑟𝑎2 represents the second stage refractive index, also initially set to 0. The 

symbol 𝛼 (alpha) is used to denote the refractive step, which is a key factor in the light propagation strategy. Other 

parameters include FEs, which stands for the number of function evaluations, and MaxFEs, representing the maximum 

number of function evaluations allowed for the optimization process. The algorithm also utilizes a population of 

individuals, where each individual represents a potential solution in the search space. The quality of these individuals is 

assessed using a fitness value, with the best and worst fitness values denoted as 𝑓𝑖𝑡𝑏𝑒𝑠𝑡 and 𝑓𝑖𝑡𝑤𝑜𝑟𝑠𝑡, respectively. The 

algorithm iteratively updates these parameters to enhance the search efficiency and convergence speed. 

3.3. Principle of Light Propagation 

The light propagation principle in FATA is executed after the mirage light filtering principle, which acts as an 

individual search strategy for the algorithm and is responsible for local exploitation in the search space to find local 

minima, as shown in Figure 11. 

 

Figure 11. FATA algorithm light propagation strategy 

The specific formula for light refraction (first stage) is as follows: 

𝑥𝑛𝑒𝑥𝑡 = 𝑥𝑏𝑒𝑠𝑡 + 𝑥𝑧  (5) 

𝑥𝑧 = 𝑥 ⋅ 𝑃𝑎𝑟𝑎1  (6) 

𝑃𝑎𝑟𝑎1 =
𝑠𝑖𝑛(𝑖1)

𝐶⋅𝑐𝑜𝑠(𝑖2)
= 𝑡𝑎𝑛(𝜃)  (7) 

where, 𝑥best  denotes the optimal individual, 𝑥𝑧  denotes the refractive step, 𝑃 ara 𝑎1  denotes the first stage refractive 

index, 𝑖1 denotes the angle of incidence, 𝑖2 denotes the angle of refraction, and 𝜃 denotes the angular change of the 

FATA algorithm, which is shown schematically in Figure 12. 

 

Figure 12. The first stage of the refraction process 

The variation curve of the parameter 𝑃𝑎𝑟𝑎1 with the number of iterations is shown in Figure 13.  



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

870 

 

 

Figure 13. Para1 trends 

The specific formula for light refraction (second stage) is as follows: 

𝑥𝑛𝑒𝑥𝑡 = 𝑥𝑏𝑒𝑠𝑡 + 𝑥𝑠  (8) 

𝑥𝑠 = 𝑥𝑓 ⋅ 𝑃𝑎𝑟𝑎2  (9) 

𝑃𝑎𝑟𝑎2 =
𝑐𝑜𝑠(𝑖3)

𝐶⋅𝑠𝑖𝑛(𝑖4)
=

1

𝑡𝑎𝑛(𝜃)
  (10) 

where, 𝑥𝑠 represents the second stage refraction step, Para2 represents the first stage refractive index, 𝑥𝑓 represents the 

light individual and the refraction process is shown in Figure 14. 

 

Figure 14. The second stage of the refraction process 

The variation curve of the parameter Para2 with the number of iterations is shown in Figure 15. 

 

Figure 15. Para2 trends 

The total reflection model is calculated as follows: 

𝑥𝑛𝑒𝑥𝑡 = 𝑥𝑓 = 0.5 ⋅ (𝛼 + 1)(𝑈𝑏 + 𝐿𝑏) − 𝛼𝑥  (11) 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

871 

 

𝛼 =
𝐹

𝐸
  (12) 

𝑥0 − 𝑥𝑓 =
𝐹⋅(𝑥−𝑥0)

𝐸
  (13) 

𝑥0 =
𝑈𝑏−𝐿𝑏

2
+ 𝐿𝑏 =

𝑈𝑏+𝐿𝑏

2
  (14) 

where 𝑥𝑓 is the total reflection model emitting individual and 𝛼 is the reflectivity, the total reflection strategy is shown 

in Figure l6. 

 

Figure 16. Total-reflective strategy 

The pseudo-code of the FATA algorithm is shown in Table 1, and the specific flowchart is shown in Figure 17. 

Table 1. Pseudo0cond of the fata algorithm 

Algorithm 1: FATA algorithm pseudo-code 

Inputs: the FATA parameters n, d, MaxFEs; 

Output: optimal individuals for the FATA algorithm; 

1 Initialise the FATA algorithms Para1, Para2, α; 

2 Initialize the FATA algorithm population; 

3 Calculate the FATA algorithm light adaptation value; 

4 FEs=0; 

5 While FEs < MaxFEs 

6 Update the optimal solution and optimal value; 

7 Calculate the weights P; calculate the parameters Para1 and Para2; 

8 If rand>P 

9 Random initialization of light populations; 

10 Else 

11 If rand<p 

12 Update the population according to the Phase 1 refraction strategy; 

13 Else 

14 Update the population according to the second phase of the refraction strategy; 

15 Using all-reflective renewal populations; 

16 End if 

17 End if 

18 t=t+1; 

19 End while 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

872 

 

 

Figure 17. Flowchart of FATA algorithm 

4. FATA-ResNet network and CAD/CAM application 

4.1. ResNet Neural Network 

ResNet (Residual Neural Network) [26] is a deep learning model proposed by researchers at Microsoft Research in 

2015, which makes it possible to train very deep networks by introducing residual learning and solving the gradient 

vanishing problem, as illustrated in Figure 18. The core innovation of ResNet is its Residual Block design, which allows 

the network to learn the residuals between inputs and outputs through Skip Connection rather than directly learning 

complex nonlinear mappings [27]. The core innovation of ResNet is its Residual Block design, which allows the network 

to learn the residuals between inputs and outputs through Skip Connection instead of learning complex nonlinear 

mappings directly [27]. 

 

Figure 18. ResNet neural network structure 

The basic unit of ResNet is the residual block, which consists of two or more convolutional layers that are usually 

followed by Batch Normalisation and ReLU activation functions. The output of the residual block consists not only of 

the results of these convolutional layers but also of the input itself, which is realized by element-level summation [28]. 

This structure helps the gradient to flow more easily during backpropagation, thus alleviating the gradient vanishing 

problem [29]. 

To enhance the CAD/CAM integration technology to improve the analysis accuracy, this paper adopts ResNet50 

[30]. ResNet50 contains 49 convolutional layers and 1 fully connected layer. The overall structure of ResNet50 is shown 

in Figure 19. The overall structure consists of the Conv convolutional layer, the Batch Norm normalization layer, the 

ReLU activation function, and other basic structures. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

873 

 

 

Figure 19. Structure of ResNet50 

4.2. FATA-ResNet Network 

To further improve the analysis accuracy of CAD/CAM integration technology applications under the cloud 

manufacturing framework, this paper proposes the FATA-ResNet network model. The FATA algorithm enhances 

convergence and accuracy through its unique search strategies. The Mirage Light Filtering Principle (MLF) evaluates 

and filters the population based on the definite integration principle, improving exploration capability. The Light 

Propagation Strategy (LPS), leveraging trigonometric principles, accelerates convergence and boosts exploitation 

efficiency. Combined, these strategies optimize both population and individual search abilities. Comparative 

experiments with other optimization algorithms on 23 benchmark functions and IEEE CEC 2014 show FATA's superior 

optimization capability. Furthermore, when applied to three practical engineering optimization problems, FATA 

outperformed its counterparts, proving its effectiveness in real-world scenarios. The model is to construct the FATA-

ResNet network model by taking the ResNet network parameters as the optimization object and the loss function as the 

adaptation value, and the specific model structure is shown in Figure 20. 

 

Figure 20. Structure of FATA-ResNet 

4.3. CAD/CAM Integration Application Method Based on FATA-ResNet Network 

Based on cloud manufacturing, the CAD/CAM integration architecture is designed based on the demand analysis of 

CAD/CAM integration technology. The CAD/CAM integration technology under the cloud manufacturing framework 

includes two key technologies, i.e., CAD/CAM integration method and CAD/CAM integration application analysis. For 

the CAD/CAM integration application analysis problem, this paper proposes a CAD/CAM integration application 

analysis method based on the FATA-ResNet network, and its specific flow chart is shown in Figure 21. 

 

Figure 21. Flow of CAD/CAM integration application 

5. Results and Discussion 

5.1. Experimental Set-up 

Experiments were conducted on a computer with the following configurations: 1) Windows 10 operating system, 16 

GB RAM, AMD Ryzen7 7735H-3.20 GHz processor for the CPU, and RTX4060 for the GPU; 2) using the programming 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

874 

 

language Python 3.10 and the deep learning framework Pytorch 1.7.1, the experiments were set up with a training number 

of times is 40 epochs, and the AdamW optimizer is used in the training process. In the experimental setup, the dataset 

used for training and testing the FATA-ResNet model comprised 5000 samples collected from actual industrial 

CAD/CAM integration scenarios. The data were formatted as 200 - dimensional feature vectors in CSV files, with each 

vector representing a specific CAD/CAM integration case. These cases included information on 3D model complexity, 

process file parameters, and integration outcomes. The dataset was divided into a training set (70%, 3500 samples) and 

a testing set (30%, 1500 samples). Before use, data underwent normalization to a [0, 1] scale to enhance model training 

stability and efficiency. This dataset provided a realistic basis for evaluating the model's performance in practical 

applications. 

 Data collection from CAD/CAM systems involved extracting operational data and integration outcomes via 

automated scripts. This data, stored in CSV files, included 200 - dimensional feature vectors representing 3D model 

attributes and process parameters. Preprocessing steps included data cleaning to remove incomplete or inconsistent 

entries, followed by normalization using Min - Max scaling to adjust all feature values to a [0, 1] range. This ensured 

consistent input for the FATA-ResNet model, enhancing training stability and accuracy. The model was trained and 

deployed using a computer with a Windows 10 OS, 16 GB RAM, an AMD Ryzen7 7735H CPU (3.20 GHz), and an 

RTX4060 GPU. Given its moderate computational demands, the model could potentially run in factories with limited 

resources. However, deployment in such environments would require optimization to reduce memory and processing 

requirements. 

Training is performed on CAD/CAM integrated design case datasets to validate the effectiveness and stability of the 

integrated application analysis models in this study. The algorithms compared include SCA-ResNet, AOA-ResNet, 

ACOR-ResNet, DE-ResNet, and ResNet, where the parameter settings of SCA-ResNet, AOA-ResNet, ACOR-ResNet, 

DE-ResNet, and FATA-ResNet optimization algorithms are shown in Table 2.  

Table 2. Parameter settings for different optimization algorithms 

Modelling Parameterization 

SCA-ResNet A=2 

AOA-ResNet α = 5; μ = 0.5 

ACOR-ResNet k = 10; ξ = 1; q = 0.5 

DE-ResNet Crossing probability = 0.5; scaling factor = 0.5 

FATA-ResNet Para1=0; Para2=0 

5.2. Analysis of Design Effects 

The CAD/CAM integration technology designed in this paper mainly includes obtaining tasks, uploading tasks, 

logging into the cloud platform, and operating models, and the specific design effects are shown in Figures 22 to 24. 

Figure 22 gives the manufacturing cloud platform login interface. Figure 23 gives the manufacturing cloud platform 

login task management interface. Figure 24 gives the browsing and editing interface of the 3D process. As can be seen 

from Figures 22 to 24, the CAD/CAM integration method designed in this paper can meet the design requirements and 

solve the CAD/CAM integration problem. 

 

Figure 22. Manufacturing Cloud Platform Login Interface 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

875 

 

 

Figure 23. Task Management Interface 

 

Figure 24. Browsing and editing interface of CAD/CAM 3D process 

Figure 22 shows the design of the login interface of the cloud manufacturing platform, which serves as the entrance 

for users to interact with the system and reflects the security and convenience of the cloud platform. The design 

incorporates a user authentication and permission management mechanism to ensure that different users (e.g., designers, 

engineers, and administrators) can access the corresponding modules and data resources according to their roles. This 

hierarchical design of permissions can effectively protect the platform's core data security and avoid potential 

information leakage due to confusing operation permissions. In addition, the interface layout is simple and intuitive, and 

the operation guidelines are clear, which helps users quickly familiarise themselves with the system functions and 

efficiently complete the task entry operations. The login interface also supports real-time connection with the cloud 

database, which allows users to enter the task management or model editing module directly after logging in, without 

additional loading, optimizing the smoothness of operation and time cost. This interface design not only improves the 

user experience but also provides a basic guarantee for the efficient operation of subsequent functional modules, 

reflecting the balance between functionality and user-friendliness in the design of the cloud manufacturing platform. 

Figure 23 shows the design of the task management interface of the cloud manufacturing platform, which is one of 

the core functions of achieving collaborative manufacturing and task optimization management. The interface layout is 

intuitive and clear, providing several key functions such as task creation, assignment, tracking, and status monitoring. 

Users can view the priority, current progress, and completion of tasks through this interface, and make dynamic 

adjustments based on real-time updated information, such as reallocating resources or modifying task parameters. The 

task management interface adopts a modular design with a clear division of labor between different functional blocks, 

allowing users to access key information efficiently. 

In addition, the interface supports cross-platform sharing and manipulation of task data, allowing users to seamlessly 

collaborate between different devices via the cloud platform, which is particularly important in a cloud manufacturing 

environment. Task status updates are displayed with visual progress bars and identifiers, further enhancing the efficiency 

and accuracy of information acquisition. Meanwhile, the task history function provides a reference basis for subsequent 

data analysis and optimization decisions. 

From the perspective of system functions, the task management interface not only improves the transparency and 

flexibility of the manufacturing process but also provides efficient support for inter-team collaboration. Its design reflects 

the adaptability of the cloud manufacturing platform for complex task decomposition and dynamic management, laying 

a technical foundation for the efficient operation of intelligent manufacturing. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

876 

 

Figure 24 shows the design of the 3D process browsing and editing interface in the cloud manufacturing platform, 

which is designed to support the user's real-time viewing and dynamic editing functions of the CAD/CAM model. The 

interface provides a comprehensive visual presentation of the 3D model, including model rotation, scaling, and detail 

zoom operations, enabling users to analyze and optimize design details from multiple perspectives and in a holistic 

manner. In addition, the interface integrates intelligent process editing tools that support online modification of model 

parameters, adjustment of process paths, and generation of new process files, which greatly enhances operational 

flexibility and efficiency. 

The 3D process interface also features real-time updates and multi-user collaboration, with user changes instantly 

synchronized to the relevant task modules via the cloud platform, allowing team members to follow up based on the 

latest design. This collaboration mechanism is especially important for geographically distributed manufacturing teams. 

At the same time, the interface design focuses on interactivity and ease of use, with a clear layout and intuitive operation 

guidelines to reduce user learning costs and improve the operability of the system. 

From the technical realization level, the 3D process interface relies on the efficient data transfer and storage 

capabilities of cloud manufacturing to ensure smooth loading and editing of complex models. 

5.3. Analysis of Calculation 

To solve the problem of analyzing the application of CAD/CAM integration technology based on cloud 

manufacturing, this paper adopts the CAD/CAM integration software based on cloud manufacturing to collect data to 

analyze and compare ResNet, SCA-ResNet, AOA-ResNet, ACOR-ResNet, DE-ResNet, and FATA-ResNet, and the 

specific results are shown in Figure 25 and Table 3. 

 

Figure 25. Optimization iteration curves for different models 

Table 3. Performance comparison results of different models 

Analytical model Accuracy/% Recall rate/% F1 score 

ResNet 87.1 83.2 0.801 

SCA-ResNet 85.3 80.8 0.765 

AOA-ResNet 86.3 82.6 0.812 

ACOR-ResNet 88.0 85.7 0.853 

DE-ResNet 92.4 90.2 0.890 

FATA-ResNet 97.6 95.3 0.945 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

877 

 

The training loss values for different models are given in Figure 25. From Figure 25, it can be seen that the training 
of different models converges to near 0 as the number of iterations increases. The learning and training convergence 
speed of the optimized ResNet model based on the FATA algorithm is better than the other models and the convergence 

accuracy is better than the other models. Figure 25 shows the optimized iteration curves of different models during 
training for comparing the convergence speed and accuracy performance of the FATA-ResNet model with the other 
compared models such as ResNet, SCA-ResNet, AOA-ResNet, ACOR-ResNet, and DE-ResNet. From the figure, it can 
be observed that the training loss values of all the models gradually decrease with the increase in the number of iterations, 
indicating that all the models can fit the data effectively during the training process. However, there are significant 
differences in the convergence speed and final loss values of different models. 

The FATA-ResNet model has the fastest decreasing training loss values, rapidly approaching the convergence state 
at the initial stage, and has the lowest final loss values. This suggests that the FATA algorithm significantly improves 

the global search capability and local exploitation of the model by optimizing the parameters of the ResNet network, 
thus speeding up the optimization process and improving the accuracy of the results. In contrast, ResNet and other 
comparison models (e.g., SCA-ResNet, AOA-ResNet, etc.) exhibit slower convergence speeds and higher final loss 
values than FATA-ResNet, indicating that these models may suffer from local optimal point traps or insufficient 
parameter tuning during the optimization process. 

The comprehensive analysis shows that FATA-ResNet outperforms other models in terms of training efficiency and 
accuracy performance, and is suitable for application in CAD/CAM integration technology analysis tasks that require 
high optimization accuracy and efficiency. This further validates the effectiveness of the FATA algorithm in optimizing 

complex neural networks. 

The accuracy, recall, and F1 scores of different models are given in Table 3. Table 3 demonstrates the 
performance comparison of different models in the analysis of CAD/CAM integration technology applications, 
including three key metrics, namely accuracy, recall, and F1 score, which are used to evaluate the classification 

accuracy, coverage ability of positive examples, and comprehensive performance of the models, respectively. The 
experiments compare ResNet, SCA-ResNet, AOA-ResNet, ACOR-ResNet, DE-ResNet, and FATA-ResNet 
models. As can be seen from the table, the FATA-ResNet model outperforms the other compared models in all 
metrics, with an accuracy of 97.6%, a recall of 95.3%, and an F1 score of 0.945, which reflects its excellent 
performance in data-driven analysis. This indicates that the FATA optimization algorithm successfully improves 
the parameter tuning capability of the ResNet network, enabling the model to more accurately capture complex 

features in CAD/CAM integration data. 

In comparison, the performance of the traditional ResNet model is low, with accuracy and recall of 87.1% and 83.2%, 
respectively, and an F1 score of only 0.801, reflecting the inadequacy of the unoptimized model in coping with complex 
data. Other optimized models (e.g., SCA-ResNet, AOA-ResNet, ACOR-ResNet, DE-ResNet), despite improving the 

performance to a certain extent, are not up to the level of FATA-ResNet, and the gap is especially obvious in the F1 
score of the comprehensive performance index. 

The experimental results demonstrate the FATA-ResNet model's superiority in analyzing CAD/CAM integration 
technology. Its high accuracy, recall, and F1 score indicate excellent classification performance. The faster convergence 
speed and higher computational efficiency mean it can handle complex data more effectively than other models. These 
advantages make it highly applicable in the cloud manufacturing environment, providing a robust tool for analyzing 
CAD/CAM integration. The study thus offers a significant advancement in the field of intelligent manufacturing. 
Overall, the superiority of the FATA-ResNet model lies in the efficiency and adaptability of its optimization algorithm, 

which not only improves the model's ability to learn complex data features but also ensures the comprehensiveness and 
reliability of the classification results, which provides effective technical support for CAD/CAM integration technology 
in the cloud manufacturing environment. 

Compared to prior studies, our results show significant improvement. Previous models like ResNet and SCA-ResNet 
exhibited lower accuracy and slower convergence rates. For example, traditional ResNet only achieved an accuracy of 
87.1%, while our FATA-ResNet model reached 97.6%. This indicates that the integration of the FATA algorithm with 
ResNet substantially enhances performance. The difference lies in our optimized parameter tuning and the efficient 
global search capability of FATA. This advancement addresses the limitations of previous research and offers a more 

effective solution for CAD/CAM integration analysis in cloud manufacturing.  

To validate the effectiveness of the FATA optimization, ablation studies were conducted. We compared the FATA - 

ResNet model with the original ResNet model under identical experimental conditions. Results showed that FATA-
ResNet achieved an accuracy of 97.6% and a rapid convergence speed, while the original ResNet only reached 87.1% 
accuracy with slower convergence. This confirms FATA's significant enhancement to ResNet's integration analysis 
capability. The FATA algorithm enhances convergence and accuracy through its unique search strategies. The Mirage 
Light Filtering Principle (MLF) evaluates and filters the population based on the definite integration principle, improving 
exploration capability. The Light Propagation Strategy (LPS), leveraging trigonometric principles, accelerates 

convergence and boosts exploitation efficiency. Combined, these strategies optimize both population and individual 
search abilities. Comparative experiments with other optimization algorithms on 23 benchmark functions and IEEE CEC 
2014 show FATA's superior optimization capability. Furthermore, when applied to three practical engineering 
optimization problems, FATA outperformed its counterparts, proving its effectiveness in real-world scenarios. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

878 

 

6. Conclusion 

Aiming at the problem of CAD/CAM technology integration application under the cloud manufacturing framework, 

this paper designs a CAD/CAM technology integration scheme for mechanical engineering under the cloud 

manufacturing framework and proposes a CAD/CAM technology integration application analysis method based on the 

FATA-ResNet model. By analyzing the demand for CAD/CAM technology integration under the cloud manufacturing 

framework, a CAD/CAM technology integration application analysis model based on FATA-ResNet is proposed, and 

data analysis is performed by using CAD/CAM technology integration design software to collect data to verify the 

FATA-ResNet model performance. ResNet model performance. The results show that the FATA-ResNet model 

accuracy, recall, and F1 score are better than other models, and the convergence speed is faster than other optimization 

algorithms. A variety of mechanical engineering CAD/CAM designs will be carried out in the future to further validate 

the performance of the FATA-ResNet model. 

Although this paper proposes a CAD/CAM integration application analysis method based on the FATA-ResNet 

network and verifies its superior performance under the cloud manufacturing framework, there are still some 

shortcomings and room for improvement. On the one hand, the dataset of the current study is mainly derived from a 

single experimental environment, which does not fully consider the diverse needs in actual complex manufacturing 

scenarios, such as cross-platform data compatibility, dynamic task allocation, etc., which may affect the wide 

applicability of the model. On the other hand, although model optimization has achieved significant results in terms of 

accuracy and efficiency, its high dependence on computational resources may limit its ability to be deployed in resource-

limited enterprises. Future research can be carried out in the following aspects: first, introducing heterogeneous data 

from multiple sources to verify the applicability of the method in complex industrial scenarios; second, optimizing the 

model structure to reduce the computational complexity and improve the efficiency of resource utilization; third, 

exploring the real-time cloud collaboration capability to support multi-user, multi-task parallel operation, providing a 

more efficient and flexible technological solution in the field of smart manufacturing. Meanwhile, the research can be 

extended to more engineering application fields to further promote the digital and intelligent development of 

manufacturing. 

7. Declarations  

7.1. Data Availability Statement 

The data presented in this study are available on request from the corresponding author. 

7.2. Funding 

The author received no financial support for the research, authorship, and/or publication of this article. 

7.3. Institutional Review Board Statement 

Not applicable. 

7.4. Informed Consent Statement 

Not applicable. 

7.5. Declaration of Competing Interest 

The author declares that they have no known competing financial interests or personal relationships that could have 

appeared to influence the work reported in this paper. 

8. References 

[1] Do, T. K. L., Chau, T. T., Vu, T. K. L., & Nguyen, T. D. (2022). Study on Calculating, Designing and Manufacturing  the Smart 

Infrared Drying System. Journal of Technical Education Science, 73, 64–73. doi:10.54644/jte.73.2022.1295. 

[2] Shukla, M., & Shankar, R. (2024). Impact Assessment of Smart Manufacturing System Implementation in Small and Medium 

Enterprises: Moderating Role of Enabling Technology and Government Support. Global Journal of Flexible Systems 

Management, 25(3), 533–557. doi:10.1007/s40171-024-00400-4. 

[3] Huo, X., & Wang, X. (2023). Internet of things for smart manufacturing based on advanced encryption standard (AES) algorithm 

with chaotic system. Results in Engineering, 20. doi:10.1016/j.rineng.2023.101589. 

[4] Bao, Y., Zhang, X., Wang, C., & Ming, X. (2024). Further expansion from smart manufacturing system (SMS) to social smart 

manufacturing system (SSMS) based on industrial internet. Computers & Industrial Engineering, 191. doi:10.1016/j.cie.2024.110119. 

[5] Corbett, C. J. (2024). OM Forum—The Operations of Well-Being: An Operational Take on Happiness, Equity, and Sustainability. 

Manufacturing & Service Operations Management, 26(2), 409–430. doi:10.1287/msom.2022.0521. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

879 

 

[6] Rudel, V., Vinogradov, G., Ganser, P., Bergs, T., Vahl, C., Frings, M., König, V., Schambach, M., Dietzel, S., & Königs, M. 

(2024). Integrating Cloud Computing, Bayesian Optimization, and Neural-Additive Modeling for Enhanced CAM Systems in 5-

Axis Milling. Procedia CIRP, 128, 532–537. doi:10.1016/j.procir.2024.04.015. 

[7] Uzun, İ., Timur, A. H., & Şenel, K. (2024). In-vitro comparison of fracture resistance of CAD/CAM porcelain restorations for 

endodontically treated molars. BMC Oral Health, 24(1). doi:10.1186/s12903-024-04983-3. 

[8] Al-Kharaz, A. A., Alwahhab, A. B. A., & Sabeeh, V. (2024). Innovative Date Fruit Classifier Based on Scatter Wavelet and 

Stacking Ensemble. HighTech and Innovation Journal, 5(2), 361–381. doi:10.28991/HIJ-2024-05-02-010. 

[9] Jha, N. K., Jasti, N. V. K., Chaganti, P. K., Kota, S., & Vijayvargy, L. (2022). Validity and reliability of sustainable supply chain 

management frameworks in Indian smart manufacturing industries. Management of Environmental Quality: An International 

Journal, 34(4), 865–901. doi:10.1108/meq-04-2022-0098. 

[10] Meka, S., Dowluru, S., & Dumpala, L. (2024). Automatic Feature Recognition Techniques for the Integration of CAD and CAM: 

A Review. Smart and Sustainable Manufacturing Systems, 8(1), 83–109. doi:10.1520/SSMS20230016. 

[11] Yang, Z., Tan, Z., Zhen, L., Zhang, N., Liu, L., & Fan, T. (2024). Column generation for service assignment in cloud-based 

manufacturing. Computers &amp; Operations Research, 161, 106436. doi:10.1016/j.cor.2023.106436. 

[12] Qi, A., Zhao, D., Heidari, A. A., Liu, L., Chen, Y., & Chen, H. (2024). FATA: An efficient optimization method based on 

geophysics. Neurocomputing, 607, 128289. doi:10.1016/j.neucom.2024.128289. 

[13] Ye, L. (2024). Design and manufacturing of mechanical parts based on CAD and CAM technology. Engineering Research 

Express, 6(4), 45411. doi:10.1088/2631-8695/ad8056. 

[14] Wei, J., Zhang, X., Ji, Z., Wei, Z., & Li, J. (2022). DPLRS: Distributed Population Learning Rate Schedule. Future Generation 

Computer Systems, 132, 40–50. doi:10.1016/j.future.2022.02.001. 

[15] Eid, N. K. (2025). A Review on the Power of CAD/CAM Technology and the Material Science in Modern Manufacturing. ERU 

Research Journal, 4(1), 2223–2250. doi:10.21608/erurj.2025.299610.1167. 

[16] Hewa, T., Braeken, A., Liyanage, M., & Ylianttila, M. (2022). Fog Computing and Blockchain-Based Security Service 

Architecture for 5G Industrial IoT-Enabled Cloud Manufacturing. IEEE Transactions on Industrial Informatics, 18(10), 7174–

7185. doi:10.1109/tii.2022.3140792. 

[17] Wu, D., Rosen, D. W., Wang, L., & Schaefer, D. (2015). Cloud-based design and manufacturing: A new paradigm in digital 

manufacturing and design innovation. Computer-aided Design, 59, 1-14. doi:10.1016/j.cad.2014.07.006. 

[18] Cheng, X., Liu, Z., & Ning, Y. (2022). Editorial: Security of cloud service for the manufacturing industry. Transactions on 

Emerging Telecommunications Technologies, 33(4), 33. doi:10.1002/ett.4369. 

[19] Chang, R. I., Lin, J. Y., & Hung, Y. H. (2024). Cloud-Based Machine Learning Methods for Parameter Prediction in Textile 

Manufacturing. Sensors, 24(4), 1304. doi:10.3390/s24041304. 

[20] Tian, S., Xie, X., Xu, W., Liu, J., & Zhang, X. (2021). Dynamic assessment of sustainable manufacturing capability based on 

correlation relationship for industrial cloud robotics. The International Journal of Advanced Manufacturing Technology, 124(9), 

3113–3135. doi:10.1007/s00170-021-08024-z. 

[21] Popchev, I., Radeva, I., & Doukovska, L. (2023). Oracles Integration in Blockchain-Based Platform for Smart Crop Production 

Data Exchange. Electronics (Switzerland), 12(10), 2244. doi:10.3390/electronics12102244. 

[22] Yin, X., Zhao, Z., & Yang, W. (2023). Optimizing cleaner productions of sustainable energies: A co-design framework for 

complementary operations of offshore wind and pumped hydro-storages. Journal of Cleaner Production, 396. 

doi:10.1016/j.jclepro.2022.135832. 

[23] Ali, M. I., Lai, N. S., & Abdulla, R. (2024). Predictive maintenance of rotational machinery using deep learning. International 

Journal of Electrical and Computer Engineering, 14(1), 1112–1121. doi:10.11591/ijece.v14i1.pp1112-1121. 

[24] Ghaly, M., Elbeltagi, E., Elsmadony, A., & Tantawy, M. A. (2024). Integration of Blockchain-Enabled Smart Contracts in 

Construction: SWOT Framework and Social Network Analysis. Civil Engineering Journal, 10(5), 1662–1697. 

doi:10.28991/CEJ-2024-010-05-020. 

[25] Fernandes, G. V. O., & NassaniNassani, L. M. (2023). The Implementation of CAD/CAM Technology at Schools of Dentistry: 

A Short Communication. ENVIRO Dental Journal, 5(1), 06–08. doi:10.12944/edj.05.01.03. 

[26] Shan, L., & Zhang, L. (2022). Application of Intelligent Technology in Facade Style Recognition of Harbin Modern Architecture. 

Sustainability (Switzerland), 14(12), 7073. doi:10.3390/su14127073. 

[27] Maciura, Ł., Cieplak, T., Pliszczuk, D., Maj, M., & Rymarczyk, T. (2023). Autonomous Face Classification Online Self-Training 

System Using Pretrained ResNet50 and Multinomial Naïve Bayes. Sensors, 23(12), 5554. doi:10.3390/s23125554. 



HighTech and Innovation Journal         Vol. 6, No. 3, September, 2025 

880 

 

[28] Ben Nasr Barber, F., & Elloumi Oueslati, A. (2024). Human exons and introns classification using pre-trained Resnet-50 and 

GoogleNet models and 13-layers CNN model. Journal of Genetic Engineering and Biotechnology, 22(1), 100359. 

doi:10.1016/j.jgeb.2024.100359. 

[29] Chen, M., Niu, R., & Zheng, W. (2022). Adaptive multi-scale neural network with Resnet blocks for solving partial differential 

equations. Nonlinear Dynamics, 111(7), 6499–6518. doi:10.1007/s11071-022-08161-4. 

[30] Haruna, U., Ali, R., & Man, M. (2023). A new modification CNN using VGG19 and ResNet50V2 for classification of COVID-

19 from X-ray radiograph images. Indonesian Journal of Electrical Engineering and Computer Science, 31(1), 369–377. 

doi:10.11591/ijeecs.v31.i1.pp369-377.