Computer vision-based smart food drying method with temperature and air speed control

Abstract

The invention relates to a method that enables autonomous control of food- drying processes using renewable energy sources and optimizes energy efficiency and product quality throughout the process. In this method, deep learning-based artificial intelligence algorithms are used in conjunction with computer vision to dynamically adjust the temperature and airflow for the product being dried, thereby minimizing energy consumption throughout the process while maintaining optimal product quality. Furthermore, the innovative explainable artificial intelligence model automatically reveals understandable, optimization-based rules for which input data and related ranges determine the drying performance (energy consumption and drying efficiency) value ranges.

Description

[0001] DESCRIPTION

[0002] COMPUTER VISION-BASED SMART FOOD DRYING METHOD WITH TEMPERATURE AND AIR SPEED CONTROL

[0003] TECHNICAL FIELD

[0004] The invention relates to a method that enables autonomous control of fooddrying processes using renewable energy sources and optimizes energy efficiency and product quality throughout the process. In this method, deep learning-based artificial intelligence algorithms dynamically adjust the temperature and air speed of the dried product using computer vision, thereby minimizing energy consumption throughout the process while maintaining optimal product quality. Furthermore, the innovative explainable artificial intelligence model automatically reveals explanatory, understandable optimization-based rules regarding which input data and their related ranges are linked to the drying performance (energy consumption and drying efficiency) value ranges.

[0005] PRIOR ART

[0006] Food drying is a process applied to extend the shelf life of food products while preserving their nutritional value. Energy efficiency and product quality are among the most significant challenges encountered in the food drying industry. Many conventional drying methods fail to strike a balance between energy efficiency and product quality.

[0007] Increasing energy consumption costs in traditional drying systems raise expenses, while maintaining constant temperature and air speed during the drying process can lead to losses in the nutritional value and quality of the dried products. Failure to adjust the appropriate temperature and air speed for each product in the drying process can cause deterioration in the texture, color, and overall quality characteristics of the products. Furthermore, the intensive energy use of drying processes creates a need for sustainable solutions.

[0008] Most artificial intelligence- and machine learning-based studies used in food drying processes rely on "black box" models, which limit transparency and explainability of results. In these systems, the model's decision mechanisms and the effects of input parameters on results are unclear. This situation makes it difficult to understand the accuracy and reliability of decisions made during the optimization of drying processes. Furthermore, there is insufficient information on how users can develop or adapt these models.

[0009] In the known technique, the search space is modified by applying operations such as quantization, discretization, and fuzzification on the data. Such operations may limit the model's search area, leading to the neglect or misinterpretation of specific important parameters. This situation has been observed to negatively impact both process accuracy and the quality and energy efficiency of the results.

[0010] In current systems, the effects of important parameters, such as product quality and energy consumption, on each other have not been adequately accounted for, and real-time monitoring of these parameters has not been achieved. This situation reduces energy efficiency and negatively affects the accuracy of the drying process. Open-loop drying systems cause the product to dry later than closed-loop systems, thereby increasing the time required and the overall cost.

[0011] Traditional drying systems typically operate via an external computer and therefore have limitations in terms of portability and ergonomics. These systems, which require an external connection, cannot perform image processing efficiently, which creates difficulties for practical applications. The lack of portable, autonomous, and fully integrated methods causes functional difficulties and additional costs in drying processes. In this context, there is a need for sustainable, innovative methods that optimize energy efficiency and product quality.

[0012] BRIEF DESCRIPTION OF THE INVENTION

[0013] The invention relates to a method that enables autonomous control of fooddrying processes using renewable energy sources and optimizes energy efficiency and product quality throughout the process. This method improves both energy consumption and the quality of the dried product in food drying processes by utilizing an explainable artificial intelligence-based control system and deep learning-based computer vision algorithms. This innovative method, which dynamically adjusts temperature and air speed through closed-loop humidity control based on the external structure of food products and utilizes solar energy as an alternative energy source, offers a more sustainable, faster solution for drying products in the food industry. Furthermore, our invention, which operates with solar energy support from renewable sources and can autonomously adjust temperature and airflow, is used for the drying of agricultural products. Our invention provides dynamic control of the interaction between drying air and the product structure using computer vision and machine learning algorithms, achieving optimal energy consumption and product quality. This method focuses on energy efficiency while also accelerating the drying process and preserving the product's nutritional value. In this context, the invention contributes to public health by ensuring that dried foods reach consumers without losing their nutritional value. Furthermore, an energy-efficient food drying method is expected to help reduce the process's carbon footprint.

[0014] The advantages of our invention are listed below:

[0015] - It provides lower energy costs and a sustainable drying process compared to traditional drying methods.

[0016] - Equipped with computer vision and machine learning-based algorithms, it autonomously adjusts temperature and air speed during the drying process, ensuring optimal drying conditions throughout.

[0017] - It preserves the color, texture, and nutritional value of products, resulting in high-quality dried products.

[0018] - By performing the drying process at different air speeds with closed- loop humidity control, it provides a faster drying time compared to traditional methods, thereby increasing production capacity.

[0019] - It ensures environmental sustainability with its structure powered by solar energy, a renewable energy source.

[0020] - It can support various drying scenarios with its structure that can be easily adapted to different types of products.

[0021] LIST OF FIGURES

[0022] Figure 1. Isometric View of the System

[0023] Correspondences of Numbers in the Figures

[0024] 1. Drying cabinet

[0025] 2. Fresh air inlet flap and motor

[0026] 3. Air speed sensors

[0027] 4. Temperature-humidity sensors

[0028] 5. Heater 6. Exhaust flap and motor

[0029] 7. Drying air channel

[0030] 8. Axial fan

[0031] 9. Drying tray

[0032] 10. Laser temperature sensors

[0033] 11. Load cells

[0034] 12. 3D depth camera

[0035] 13. Solar energy system

[0036] 14. PLC automation unit

[0037] 15. LED lighting

[0038] 16. GPU-based microprocessor card

[0039] DETAILED DESCRIPTION OF THE INVENTION

[0040] Invention: drying cabinet (1 ), fresh air inlet damper and motor (2), air velocity sensors (3), temperature-humidity sensors (4), heater (5), exhaust damper and motor (6), drying air duct (7), axial fan (8), drying tray (9), laser temperature sensors (10), load cells (11 ), 3D depth camera (12), solar energy system (13), PLC automation unit (14), LED lighting (15), and GPU-based microprocessor card (16).

[0041] The air entering the drying cabinet (1 ) through the fresh air inlet flap and motor (2) first passes through the air velocity sensors (3) and temperature-humidity sensors (4) before reaching the product being dried. The heated, drying air from the heater (5) then absorbs moisture from the product, facilitating heat and mass transfer.

[0042] When the relative humidity values of the air entering and exiting the drying cabinet (1 ) are equal, the fresh air inlet flap and motor (2) and the exhaust flap and motor (6) are opened, allowing the moisture-saturated air in the drying air duct (7) to be removed from the system via the axial fan (8) and fresh air to be introduced into the system .

[0043] In our invention, a total of 9 parameters are measured: drying chamber inlet temperature, drying chamber inlet humidity, product surface temperature, air velocity over the product, weight change of the product placed on the drying tray (9) during drying, drying chamber outlet temperature, drying chamber outlet humidity, total energy consumption, and product images. The temperature-humidity values at the drying cabinet inlet and outlet are measured using temperature-humidity sensors (4), the product surface temperature is measured using laser temperature sensors (10), the air velocity passing over the product is measured with an air velocity sensor (3), the weight change in the product is measured with load cells (11 ), and product images are obtained with a two-dimensional and three-dimensional 3D depth camera (12). Additionally, the system includes two inverter drivers that adjust the heater (5) power and axial fan (8) speed, one data acquisition system to receive data from the sensors, and two power meters connected to the heater (5) and axial fan (8) elements to measure energy consumption.

[0044] The energy for the heater (5) and axial fan (8) is provided by a solar energy system (13) that includes a gel battery integrated into the photovoltaic solar panel, making the energy sustainable. A Programmable Logic Controller (PLC) automation unit (14) is also used to control the system.

[0045] The food drying process in our invention occurs in two distinct stages: pre-drying and final drying. During the pre-drying stage, the system optimizes energy consumption, generates transparent artificial intelligence models, and determines drying scenarios that will ensure optimal energy consumption and drying quality based on model outputs. In the final drying stage, a deep learning-based computer vision system is used to control product quality, and the drying process is controlled based on image data.

[0046] In the computer vision system, images captured by the camera are processed at intervals specified by the user to identify product color, surface, and texture during the drying process. The data is also transmitted in real time to the drying system's control panel.

[0047] To capture images of the products on the drying tray (9) using the 3D depth camera (12) in a well-lit environment, LED lighting (15) is installed in the drying cabinet (1 ), and the images are processed on the Graphics Processing Unit (GPU)-based microprocessor card (16). Stereographic images are also captured at specific time intervals using the depth sensor of the 3D depth camera (12), and the product's surface height is measured to support image processing for product quality assessment. Thus, with our invention, an intelligent drying system with image processing capabilities determines optimal drying conditions that ensure product quality and system energy efficiency, enabling faster drying with closed-loop humidity control.

Claims

CLAIMS1. A computer vision-based smart food drying method that utilizes renewable energy sources to control temperature and drying air speed, characterized by;- fresh air enters the drying chamber (1) through the air inlet flap, and the motor (2) passes through air velocity sensors (3) and temperaturehumidity sensors (4) before reaching the product being dried,- the drying air, heated by the heater (5), then absorbs the moisture on the product, performing heat and mass transfer,- the relative humidity values of the air entering and exiting the drying cabinet (1 ) equal, the fresh air inlet flap and motor (2) and the exhaust flap and motor (6) opened to remove the moisture-saturated air in the drying air duct (7) from the system using an axial fan (8) and to bring fresh air into the system,- measuring the drying cabinet inlet temperature, drying cabinet inlet humidity, product surface temperature, air velocity over the product, weight change of the product placed on the drying tray (9) during drying, drying cabinet outlet temperature, drying cabinet outlet humidity, total energy consumption, and product images,- food drying in two different stages: pre-drying, which uses a deep learning-based computer vision system to optimize system energy, produce transparent artificial intelligence models, and determine drying scenarios based on model outputs, and final drying, which controls the drying process based on image data,- installation of LED lighting (15) in the drying cabinet (1 ) to enable images of products on the drying tray (9) to be captured in a well-lit environment using a 3D depth camera (12),- image processing on a GPU-based microprocessor card (16).

Images