Automating the Art of Jalebi Manufacturing: A Machine Learning-Assisted Inverse Kinematics Approach for Precision Path Planning and Scalable Production of a Traditional Indian Sweet
MaheshA. Makwana1✉Emailmaheshmakwana@aau.in
SamitDutta2Emailsamitdutta@aau.in
KalpeshParmar3Emailkalpesh.polytech@gmail.com
1
A
Food Process Engineering DepartmentCollege of Food Processing Technology and BioenergyAnand Agricultural University388110AnandGujaratIndia2Food Business Management DepartmentCollege of Food Processing Technology and BioenergyAnand Agricultural University388110AnandGujaratIndia
3Department of Electronics and CommunicationGovernment Polytechnic CollegeSector 30382028GandhinagarGujaratIndia
A
Abstract
Jalebi
is a popular deep-fried Indian dessert made from fermented batter, typically consisting of refined flour and yogurt. The automation of
jalebi
making is essential to improve efficiency, consistency, and hygiene in large-scale production. Traditional manual preparation lacks hygiene, requires intensive labor, consumes time, and results in inconsistencies in shape, size, and texture. An automation in jalebi-making ensures uniform batter dispensing, precise frying, and controlled sugar syrup absorption, reducing human effort and production costs. Additionally, automation enhances food safety by minimizing direct human contact, making it an ideal solution for commercial kitchens, sweet shops, and industrial-scale food production. The automation of
Jalebi
production presents a significant challenge due to intricate motion patterns traditionally performed by skilled artisans. Conventional inverse kinematics approaches struggle to replicate these patterns due to the absence of a mechanistic motion model. This study proposes a novel machine learning-based path planning method for the automated production of
Jalebi
, ensuring consistent shape, quality, and texture while preserving traditional craftsmanship. Motion data was collected from skilled artisans to capture path patterns, size variations, and production speeds. A parametric equation was developed to represent the
Jalebi
formation process, enabling the formulation of an inverse kinematics solution for an RUU-parallel robot configuration. To the best of our knowledge, this is the first attempt at automating
Jalebi
making using an RUU-parallel manipulator, a low cost solution. A supervised learning model was trained on these motion patterns, achieving a mean squared error of
after 1000 iterations and an R-squared value of 1. The trained network was validated which shows the mean square error of 0.0988 and an optimized R-squared value 0.8742. The approach was validated through simulations, demonstrating accuracy, consistency, and adaptability across different
Jalebi
designs and production scales. This framework enables scalable, precise, and efficient
Jalebi
manufacturing, with potential applications in automating other intricate food production processes, such as Chakari and Imarti, thereby bridging cultural heritage with modern automation technologies. In addition, the conceptualized Automatic Jalebi Making Machine not only optimizes the automation process but also provides a cost-effective and scalable solution for similar food preparation industries.
Keywords
Inverse Kinematics
Jalebi
Path Planning
Automation
Parallel Manipulator
A
A
A
A
A
Introduction
The integration of artificial intelligence (AI) and machine learning (ML) into traditional food production processes has been transformative, improving efficiency and consistency across various culinary practices \citep{caldwell2023automation}. These technologies can enhance production processes, improve supply chain management, and ensure food safety \citep{kler2022retracted}. They also enhance food quality, hygiene, productivity, and workplace safety \citep{barasa2023robotics}. Recent advancements include novel sensors, end effectors, and robotic cutting, with potential applications in cleaning, mixing, dough manipulation, precision dosing/cooking, and additive manufacturing \citep{derossi2023avenues}. Innovations in automation, control strategy, 3-D printing and sustainable practices have improved product scalability and marketability cheruvu2008recent,makwana2021forward. The food industry has adopted sophisticated automation, control, and monitoring methods, including robotics and modular techniques \citep{james2005food}.
Novel technology such as three-dimensional (3-D) food printing is an emerging technology revolutionizing the food industry by enabling personalized, complex food products with tailored shapes, flavors, textures, and nutritional content \citep{hamilton2024bytes}. The technology employs layer-by-layer deposition to create intricate food structures, with various printing methods, platforms, materials, and recipes being utilized sun20183d. Additionally, this technology shows promise in automating food production, potentially reducing food waste and improving efficiency \citep{leontiou2023three}. As research progresses, 3-D food printing is poised to transform food manufacturing, offering personalized nutrition and innovative product designs. A prototype 3-D printer based on a delta parallel robot demonstrated decreased printing time without compromising quality \citep{celi2015study}. Another study developed a dual-extruder food 3-D printer using a Selective Compliance Assembly Robot Arm (SCARA), which improved productivity and showed broad adaptability for various food inks \citep{pan2024design}. These advances in 3-D food printing are paving the way for innovative food production approaches and localized manufacturing.
A
Recent research has explored innovative approaches to automating and enhancing traditional Indian desserts, particularly jalebi. The Data Jalebi Bot project utilizes custom software to generate edible data visualizations in the form of jalebi, transforming professional profile summaries into consumable data sculptures \citep{patekar2018data}. Despite these advancements, automating the production of traditional desserts that require intricate handcrafting remains a formidable challenge \citep{patoliya2023conceptual}. Jalebi, a renowned Indian sweet as shown in figure 2 known for its complex spiral shapes, exemplifies this difficulty. The traditional method relies heavily on the dexterity and experience of artisans to achieve its characteristic form and texture. Replicating these precise motion patterns as shown in figure 3 through automation requires advanced control systems capable of mimicking human hand movements \citep{choudhury2024socioeconomic}.Fig. 3
Cooking of Jalebi
Conventional mechanistic models often fall short in accurately capturing the nuances of Jalebi formation, making it arduous to solve inverse kinematics for automated systems\citep{de2018inverse}. This limitation underscores the need for data-driven approaches that can learn and replicate the intricate trajectories involved in the process \citep{gholami2021inverse} . Recent developments have seen the application of ML in agriculture to optimize complex tasks, suggesting potential applicability in food production \citep{jha2019comprehensive}.
This study introduces a novel machine learning-based path planning framework aimed at automating Jalebi production utilizing 3-D printing approach with delta robot. By analyzing motion data from skilled artisans, we developed a parametric model to represent the formation process. This model facilitated the creation of an inverse kinematics solution tailored for an RUU-parallel robot configuration, enabling the precise replication of traditional Jalebi spirals. A supervised learning algorithm was trained to emulate these patterns, achieving high accuracy in reproducing the desired shapes. The proposed system was validated through simulations, demonstrating its potential to maintain the authentic craftsmanship of Jalebi making while enhancing production efficiency.
Materials and Methods
A
A brief market survey was conducted to analyze the motion patterns employed by traditional Jalebi makers. This study aimed to capture the intricate hand movements and techniques used in shaping Jalebis. Based on the findings of this survey, the standard diameter for a single Jalebi was determined to be 80 mm, ensuring consistency in size across automated production.A
The observed motion pattern consists of a central spiral surrounded by six additional spirals, forming the complete structure of a Jalebi. This pattern, as illustrated in figure 5 and figure 6, serves as the foundation for developing an optimized path planning strategy. By accurately replicating these handcrafted movement, the automation process can achieve high precision in shape formation while preserving the authenticity of traditional Jalebi-making techniques.Fig. 6
3-D continuous curvature of motion pattern of Jalebi
Mathematical Model of Motion Pattern
Based on the studied motion patterns, a mathematical model 1 was developed to accurately represent the path required for Jalebi formation.These equations form a parametric spiral in the horizontal plane (X-Y), with a constant vertical height (Z), well-suited for layering batter along a spiral route. This model serves as the foundation for determining the precise movements necessary to replicate the handcrafted spiral patterns.
eq:01
Geometric Interpretation and properties
The terms
and
suggest a circular motion, scaled by the factor
, which is a linearly increasing function of angular parameter.
This implies that the motion follows a radial expansion from the origin, meaning the path is a spiral in the
-plane with a constant
-height. Equation 2 represents the implicit form of the parametric equation in the
-plane.
The equation for
is constant, indicating a horizontal spiral at
.
Thus, the motion describes a spiral in the
-plane, expanding outward as
increases.
Continuity
Geometric Continuity: The functions used (
,
, and the linear terms) are smooth and differentiable for all real values of
, ensuring that the robot’s motion remains continuous. There are no abrupt changes in position, direction, or speed within the designed time range.\\
Path Consistency: \\The absence of breaks along the spiral path makes it ideal for robotic execution, minimizing stress on hardware and producing visually flawless Jalebi shape. \\
Precision: \\The very small constant (
) added to
is effectively zero for practical purposes and is included for numerical stability.
Implicit Representation of Parametric Equation
Implicit equations allow direct differentiation using implicit differentiation, making it easier to analyze slopes, curvature, and other geometric properties.
eq:02
A
To implement this motion in an automated system, Artificial Neural Network (ANN) as shown in figure 8 based inverse kinematics was employed to compute the joint angles required for the three actuators of an RUU-parallel manipulator. Fig. 9
RUU-parallel manipulator
Using the mathematical model developed, the inverse kinematics algorithm as attached in appendix \ref{appendix:IKcode} calculates the necessary joint positions as output, to achieve the desired end-effector path act as input. This process ensures that the parallel manipulator can precisely follow the predefined spiral motion while maintaining consistency in Jalebi formation. The kinematic configuration of the RUU-parallel manipulator used for this purpose is illustrated in figure 9.
Artificial Neural Network-Based Inverse Kinematics Solution for the RUU Parallel Manipulator
Inverse kinematics (IK) is a fundamental problem in robotics, particularly for parallel manipulators, where determining joint angles for a given end-effector path is highly complex due to nonlinearities and constraints. In this study, an ANN-based approach is employed to solve the inverse kinematics problem for an RUU parallel manipulator.
Training the Artificial Neural Network
To effectively map the end-effector path to the required joint angles, a supervised learning approach is used. The training dataset consists of X, Y, and Z coordinates extracted from the developed parametric equation that represents the continuous spiral pattern of Jalebi formation. These Cartesian coordinates serve as the input to the ANN, while the corresponding joint angles
of the RUU parallel manipulator form the output.
eq:03
where:
are the input features (
)
are the weights connecting input neurons to hidden neurons
are the biases for the hidden layer (
)
is the activation function in the hidden layer \textbf(logsig)
are the weights connecting hidden neurons to the output neuron
is the bias in the output layer
is the activation function in the output layer \textbf(purlin)
is the final output of the neural network
A
The hyper-parameters selected for training of ANN is shown in Table 1. the mathematical representation of neural network is given as equation 3.begin{table*}[htbp] \centering \caption{Hyperparameters for training} \vspace{10pt} \begin{tabular}{cccc}\hline Unit & Initial Value & Stopped Value & Target Value \hline Performance &
&
&
\\ Gradient &
&
&
Mu &
&
&
\\ Sum Squared &
&
&
\hline \end{tabular} \label{tab:1}\end{table*}
The neural network comprises 4 hidden layers and having 21, 15, 9 and 3 nodes in each layer, respectively. The ANN is trained using the Bayesian Regularization (BR) algorithm, a well-suited method for nonlinear regression problems. Bayesian Regularization helps improve generalization by preventing over-fitting and ensuring smooth convergence, making it particularly effective for robotics applications where precision and stability are required. Bayesian regularization is incorporated within the Levenberg-Marquardt algorithm to optimize the training process. Backpropagation is used to compute the Jacobian matrix
of the performance metric with respect to the weight and bias variables
. Each variable is updated according to the Levenberg-Marquardt algorithm as follows:
eq:JJ
eq:JE
eq:dX
where:
represents all errors,
is the identity matrix,
is the damping factor used in the Levenberg-Marquardt optimization.
Validation Methodology for Developed ANN
A
To validate the path generated by the ANN, a model-based RUU-parallel manipulator was developed within the Simulink environment as shown in 10. The manipulator's design and dynamic parameters were defined based on the in-house developed manipulator having specifications provided in Table 2. The simulation was then executed for a total duration of 70 s, during which the system computed the manipulator's motion path in cartesian space. This simulation aimed to ensure that the ANN-generated path could be accurately followed by the manipulator while maintaining smooth and precise motion. The results obtained from the simulation were analyzed to assess the accuracy and feasibility of implementing the computed trajectories in a real-world setup, ensuring optimal motion planning and control.A
Fig. 10
begin{table*}[htbp] \centering \caption{Dimentions of RUU-parallel manipulator} \vspace{10pt} \begin{tabular}{ccc} \hline Sr No. & Description & Dimensions (in mm) \\ \hline 1 & Bicep length & 150\\ 2 & Forearm parallelogram length & 380\\ 3 & Base platform radius & 147.2\\ 4 & Moving platform radius & 73.6\\ \hline \end{tabular} \label{tab:2}\end{table*}
Results and Discussions
A
The ANN trained using Bayesian Regularization achieves a best training performance of at \text{epoch 1000} as shown in figure 12, indicating an extremely low mean squared error (MSE). Initially, the MSE starts at approximately \text{
} and rapidly decreases within the first \text{400 epochs}, followed by a gradual decline below \text{
} after \text{700 epochs}. The close alignment between training (dotted blue) and testing (solid red) error curves confirms effective generalization without overfitting. The network successfully maps extracted \text{(x, y, z)} values from the developed parametric equation of continuous Jalebi spirals to the corresponding joint angles \text{(
)} of the \text{RUU parallel manipulator}, ensuring high precision in path planning. The ANN’s convergence towards the goal, as indicated by the dotted black line, validates its robustness and accuracy, making it highly suitable for real-time motion control applications.
The training state analysis of the ANN at \text{epoch 1000} reveals a well-converged model with optimal parameter tuning. The gradient value of
indicates minimal changes in the network’s weights, confirming that the training has reached a stable state as shown in figure 13. The regularization parameter
suggests a balance between minimizing the error and preventing overfitting, ensuring robustness in generalization. The total number of parameters optimized is \text{567.6729}, contributing to the model’s complexity while maintaining efficiency. The sum squared parameter value of \text{1540.8869} reflects a well-regularized weight distribution, avoiding excessively large weight values. Additionally, the validation checks remain \text{zero}, signifying that the training process did not trigger early stopping, reinforcing the reliability of the training procedure. These results validate the ANN’s ability to achieve precise learning and stable performance, making it suitable for path planning and control applications in the \text{RUU parallel manipulator}.
Fig. 13
Training states
A
The error histogram with 20 bins illustrates the distribution of errors between the target and output values of the ANN model. The majority of errors are concentrated around zero as shown in figure 15, indicating that the network has learned the underlying pattern effectively. The histogram exhibits a symmetrical distribution, with a peak at approximately , further validating the model's accuracy. The presence of a few outliers suggests minor deviations in certain instances but does not significantly impact overall performance. The training errors, represented in blue, dominate the distribution, while the test errors, shown in red, are minimal, indicating that the model generalizes well to unseen data. The zero-error reference line confirms that the error values are tightly clustered around zero, signifying a well-trained network with minimal bias and variance. This analysis supports the ANN’s effectiveness for precise path planning and control in the \text{RUU parallel manipulator}.
Fig. 16
Joint trajectories: predicted vs. targeted
The regression plots for both training and test datasets indicate a perfect fit with a correlation coefficient of
, signifying an exact linear relationship between predicted and actual values. The predicted values closely align with the ideal reference line
, confirming the model's accuracy. The presence of a well-fitted regression line in both cases and the absence of deviation suggest that the model generalizes well without signs of overfitting. The identical performance on training and test data ensures robustness, demonstrating that the model effectively captures underlying patterns and provides highly reliable predictions as shown in figure 16.
Validation of Inverse Kinematics
A
The inverse kinematics was solved using the developed ANN on validation samples. The ANN-based inverse kinematics results are presented in figure 18, providing joint angles for all the three actuators of the RUU parallel manipulator to achieve the desired motion pattern. The mean square error of validation is 0.0988 and optimized R-squared value is 0.8742 as extracted from dataset of figure 19.Fig. 19
Joint angles: actual versus predicted
Simulation of Motion Patterns
Finally, a Simulink model of the RUU-parallel manipulator simulates motion patterns based on predefined joint angles. The simulation results demonstrated a close approximation to the expected cartesian space path with planned path as shown in figure 20, validating the effectiveness of the developed model. Future improvements could focus on enhancing dynamic modeling, incorporating actuator constraints, and implementing real-time control strategies to further refine the system’s accuracy and efficiency.
Fig. 20
Comparison of path taken by ANN-simulation versus planned in cartesian coordinate
Conclusions
In this study, a Simulink model of the RUU-parallel manipulator was developed to simulate Jalebi motion patterns based on ANN solved inverse kinematics. The model was designed to accurately represent the kinematics and dynamics of the manipulator, ensuring precise path execution.
A
The provided joint angles obtained by solving inverse kinematics were fed into the simulation, and the resulting motion patterns were analyzed to evaluate the system's accuracy. The simulation results demonstrated a close approximation to the expected joint angle trajectories, validating the effectiveness of the developed model. The minimal deviation observed between the simulated and input joint angles indicates that the model effectively captures the manipulator's motion characteristics. This confirms the reliability of the Simulink-based approach in predicting the kinematic behavior of the RUU-parallel manipulator for automation of Jalebi manufacturing. The simulated trajectories can be directly implemented on the Automatic Jalebi Making Machine based on RUU-parallel manipulator to facilitate large-scale manufacturing. By integrating the optimized motion patterns into an automated system, the production process can achieve high precision, consistency, and efficiency. This approach ensures that the generated trajectories maintain uniformity in product shape and quality, making them suitable for mass production and scale up. Additionally, the validated trajectories can help minimize deviations, reduce operational errors, and enhance the overall performance of the manufacturing system. This machine represents a low-cost solution for automating the jalebi-making process while maintaining an accepted level of efficiency.
Ethical approval declarations (only required where applicable): Not Applicable.
begin{appendix}\section{MATLAB Script for Generating Jalebi Trajectory and Solving Inverse Kinematics}\label{secA1}\label{appendix:IKcode}\begin{lstlisting}[style=matlab,caption={MATLAB code for inverse kinematics computation of the Delta Robot for Jalebi path computation}]st = 500; % number of data pointst = linspace(0,30,st);r = 50;runtime = 0:0.04:19.99;x = 7.*cos(t).*(7.5.*t + 30);y = (7.*sin(t).*(7.5.*t + 30)) + (1.7343e-15);z = repmat(450, 1, st);pnts = [-x; -y; -z];plot3(-x, -y, -z);X = timeseries(x);Y = timeseries(y);Z = timeseries(z);xdot = diff(-x); ydot = diff(-y); zdot = diff(-z);xdot(st) = 0; ydot(st) = 0; zdot(st) = 0;Pxyz = [-x; -y; -z];Pxyzdot = [xdot; ydot; zdot];u = 73.6; v = -63.4; w = 37.3;L = 150; l2 = 380; l1 = 150;a2 = 150; a3 = 270; a1 = 30;Ra = 147.2;Rb = 73.6;R = [Rb - Ra; 0; 0];Pxyz = [x; y; z];TR1 = [cosd(a1) sind(a1) 0; -sind(a1) cosd(a1) 0; 0 0 1];TR2 = [cosd(a2) sind(a2) 0; -sind(a2) cosd(a2) 0; 0 0 1];TR3 = [cosd(a3) sind(a3) 0; -sind(a3) cosd(a3) 0; 0 0 1];B1 = (TR1 * Pxyz) + R;B2 = (TR2 * Pxyz) + R;B3 = (TR3 * Pxyz) + R;rx1 = B1(1,:); ry1 = B1(2,:); rz1 = B1(3,:);rx2 = B2(1,:); ry2 = B2(2,:); rz2 = B2(3,:);rx3 = B3(1,:); ry3 = B3(2,:); rz3 = B3(3,:);t31 = acosd((-x*sind(a1) + y*cosd(a1)) / l2);t32 = acosd((-x*sind(a2) + y*cosd(a2)) / l2);t33 = acosd((-x*sind(a3) + y*cosd(a3)) / l2);t21 = acosd((rx1.^2 + ry1.^2 + rz1.^2 - l1.^2 - l2.^2) ./ (2 .* l1 .* l2 .* sind(t31)));t22 = acosd((rx2.^2 + ry2.^2 + rz2.^2 - l1.^2 - l2.^2) ./ (2 .* l1 .* l2 .* sind(t32)));t23 = acosd((rx3.^2 + ry3.^2 + rz3.^2 - l1.^2 - l2.^2) ./ (2 .* l1 .* l2 .* sind(t33)));t11lr = atand(-((-l1.*rz1 - l2.*sind(t31).*cosd(t21).*rz1 + l2.*sind(t31).*sind(t21).*rx1) ./ ... (l1.*rx1 + l2.*sind(t31).*sind(t21).*rz1 + l2.*sind(t31).*cosd(t21).*rx1)));t11 = fliplr(t11lr);t12lr = atand(-((-l1.*rz2 - l2.*sind(t32).*cosd(t22).*rz2 + l2.*sind(t32).*sind(t22).*rx2) ./ ... (l1.*rx2 + l2.*sind(t32).*sind(t22).*rz2 + l2.*sind(t32).*cosd(t22).*rx2)));t12 = fliplr(t12lr);t13lr = atand(-((-l1.*rz3 - l2.*sind(t33).*cosd(t23).*rz3 + l2.*sind(t33).*sind(t23).*rx3) ./ ... (l1.*rx3 + l2.*sind(t33).*sind(t23).*rz3 + l2.*sind(t33).*cosd(t23).*rx3)));t13 = fliplr(t13lr);t31n = 90 - t31; t32n = 90 - t32; t33n = 90 - t33;allangles = [t11; t12; t13; t21; t22; t23; t31n; t32n; t33n];indata = [t11; t12; t13; t21; t22; t23];plot(t11,'r'); hold on;plot(t12,'b'); plot(t13,'g');title('Joint angles by IK');xlabel('Time (sec)');ylabel('Joint angles (Degrees)');legend('Joint-1','Joint-2','Joint-3');hold off;\end{lstlisting}\end{appendix}\bibliography{sn-bibliography}
References:
de Souza, Lucas B and Dalmedico, J{\^o}natas F and Kondo, Henrique S and Mendon{\c{c}}a, M{\'a}rcio and Montezuma, Marcio AF and Pocz{\k{e}}ta, Katarzyna (2018) Inverse Kinematics and Trajectory Planning Analysis of a Robotic Manipulator. International Journal of Advanced Engineering Research and Science 5(4): 207--214 AI Publications
Patoliya, Jenish and Vala, Kanaksinh and Akbari, Sanjay and Makwana, Mahesh (2023) Conceptual design of Khandvi making machine using 3D-computer aided design (CAD)(SolidWorks). International Research Journal of Engineering and Technology 10(10): 136-139 IRJET
Caldwell, Darwin G Automation in food manufacturing and processing. Springer Handbook of Automation, Springer, 2023, 949--971
Choudhury, Rupam (2024) Socioeconomic Impacts of Soft Robotic Automation on Workforce Dynamics in Artisanal Food Production: Balancing Efficiency with Traditional Craft Preservation. Journal of AI-Driven Automation, Predictive Maintenance, and Smart Technologies 9(12): 42--53
Jha, Kirtan and Doshi, Aalap and Patel, Poojan and Shah, Manan (2019) A comprehensive review on automation in agriculture using artificial intelligence. Artificial Intelligence in Agriculture 2: 1--12 Elsevier
Gholami, Akram and Homayouni, Taymaz and Ehsani, Reza and Sun, Jian-Qiao (2021) Inverse kinematic control of a delta robot using neural networks in real-time. Robotics 10(4): 115 MDPI
Makwana, Mahesh A and Patolia, Haresh P (2021) Forward kinematics of delta manipulator by novel hybrid neural network. International Journal of Mathematical, Engineering and Management Sciences 6(6): 1694 International Journal of Mathematical, Engineering and Management Sciences
A
Kler, Rajnish and Elkady, Ghada and Rane, Kantilal and Singh, Abha and Hossain, Md Shamim and Malhotra, Dheeraj and Ray, Samrat and Bhatia, Komal Kumar (2022) Machine Learning and Artificial Intelligence in the Food Industry: A Sustainable Approach. Journal of Food Quality 2022(1): 8521236 Wiley Online Library
Barasa, Samuel and Etene, Yonah (2023) Robotics in Food Manufacturing Industry in the Industry 4.0 Era. International Journal of Computer Science and Mobile Computing 12(8): 72--77
Derossi, A and Di Palma, E and Moses, JA and Santhoshkumar, P and Caporizzi, R and Severini, C (2023) Avenues for non-conventional robotics technology applications in the food industry. Food Research International 173: 113265 Elsevier
Patekar, Gaurav and Dudeja, Karan and Bablani, Himanshu and Bhaumik, Debanshu (2018) Data Jalebi Bot. ACM, 1--4, Extended Abstracts of the 2018 CHI Conference on Human Factors in Computing Systems
Cheruvu, Priyanka and Kapa, Sumaja and Mahalik, Nitaigour P (2008) Recent advances in food processing and packaging technology. International Journal of Automation and Control 2(4): 418--435 Inderscience Publishers
James, Tony (2005) Food factory in a pipe [automation in the food industry]. Manufacturing Engineer 84(4): 12--15 IET
Hamilton, Allyson N and Mirmahdi, Razieh S and Ubeyitogullari, Ali and Romana, Chetanjot K and Baum, Jamie I and Gibson, Kristen E (2024) From bytes to bites: Advancing the food industry with three-dimensional food printing. Comprehensive Reviews in Food Science and Food Safety 23(1): e13293 Wiley Online Library
Sun, Jie and Zhou, Weibiao and Huang, Dejian and Yan, Liangkun (2018) 3D food printing: Perspectives. Polymers for food applications : 725--755 Springer
Leontiou, Areti and Georgopoulos, Stavros and Karabagias, Vassilios K and Kehayias, George and Karakassides, Anastasios and Salmas, Constantinos E and Giannakas, Aris E (2023) Three-dimensional printing applications in food industry. Nanomanufacturing 3(1): 91--112 MDPI
Celi, Ricardo and Semp{\'e}rtegui, Ana and Morocho, Derlin and Loza, David and Alulema, Darwin and Proa{\ n}o, Mariela (2015) Study, design and construction of a 3D printer implemented through a delta robot. IEEE, 717--722, 2015 CHILEAN Conference on Electrical, Electronics Engineering, Information and Communication Technologies (CHILECON)
Pan, Jingfu and Chen, Xiaochen and Zhu, Yulin and Xu, Baoguo and Li, Changzhu and Khin, Myat Noe and Cui, Haiying and Lin, Lin (2024) Design and development of dual-extruder food 3D printer based on selective compliance assembly robot arm and printing of various inks. Journal of Food Engineering 370: 111973 Elsevier
Supplementary information
Not Applicable.
Additional Files
Additional file 6
Additional file 7
Additional file 10
Additional file 12
Additional file 14
Additional file 15
Additional file 18
Additional file 19
Additional file 20
Additional file 21
Additional file 23
Additional file 24
Additional file 26
Additional file 27
Additional file 29
Additional file 31
Additional file 32
Additional file 34
Additional file 35
Additional file 37
Additional file 40
A
Acknowledgement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
A
The authors would like to thank the editor and anonymous reviewers for their comments that help improve the quality of this work.Conflict of Interest Statement
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
A
A
Author Contribution
Author Contributions. First author conceptualized the research, designed the methodology, developed the theoretical framework, performed simulations, analyzed the results, and drafted the manuscript. Second Author facilitated access to research facilities, managed laboratory experiments, and assisted in formatting and refining the article. Third author analyse the data and providing assistance in writing. Authors reviewed and approved the final manuscript and agree to be accountable for its content.