Laser cutting optimization method and system for batch production of flexible touch sensors

Abstract

The invention provides a laser cutting optimization method and system for batch production of flexible touch sensors. The method comprises the following steps: acquiring characteristic data of the flexible material, and configuring an ultrafast laser cutting unit and a thermal annealing laser unit; plasmas are generated through the plasma generator, and the laser energy absorption efficiency is improved; collecting cutting process parameters, and estimating a cutting state; dynamically adjusting cutting beam, thermal annealing beam and plasma parameters; evaluating the cutting quality and selecting or reconstructing a control strategy; and simultaneous cutting and post-processing of a plurality of sensor units are realized. According to the flexible touch sensor cutting device, the multi-beam laser collaborative cutting and the plasma auxiliary technology are combined, the distributed unknown input observer and the intelligent control system are adopted, the problems of edge damage, large heat affected zone, material deformation and the like in traditional laser cutting are solved, and the cutting quality and the batch production efficiency of the flexible touch sensor are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of flexible electronic manufacturing technology, and in particular to a laser cutting optimization method and system for mass production of flexible tactile sensors. Background Art

[0002] Flexible tactile sensors are a new type of intelligent sensor capable of sensing and measuring tactile information. They are widely used in fields such as human-computer interaction, healthcare, and robotics. These sensors are lightweight, flexible, and wearable, and can mimic the way human skin senses external information such as pressure, temperature, and vibration.

[0003] Currently, common manufacturing technologies for flexible tactile sensors include screen printing and photolithography. Screen printing uses a specific template to print conductive materials onto a flexible substrate to form electrodes and sensitive layers. Photolithography, on the other hand, uses photosensitive materials to construct micro- and nanostructures on a flexible substrate through processes such as exposure and development. These technologies have made significant progress in realizing sensor functional structures.

[0004] In the manufacturing process of flexible tactile sensors, the cutting process is a critical step in determining the quality and performance of the final product. While traditional laser cutting technology can achieve high-precision cutting, it often faces issues such as edge damage, large heat-affected zones, and material deformation when processing flexible materials. These issues lead to unstable sensor edge structures, affecting sensing accuracy and resulting in high defect rates in mass production. This also limits the sensor's use in high-precision applications.

[0005] The main technical defects of traditional laser cutting technology include: a single laser beam cannot take into account both cutting efficiency and edge quality at the same time; the heat generated during the cutting process is difficult to effectively control, resulting in thermal deformation and microcracks of the material; stress concentration and material damage at the cutting edge are difficult to repair in real time during the cutting process; in addition, the debris and melt generated during the cutting process are easily deposited on the cutting edge again, affecting the cutting quality and subsequent processes. Summary of the Invention

[0006] The purpose of the present invention is to provide a laser cutting optimization method and system for the mass production of flexible tactile sensors, which is used to solve the technical problems faced by traditional laser cutting technology when processing flexible materials, such as edge damage, large heat-affected zone, and material deformation, and improve the quality and efficiency of the mass production of flexible tactile sensors.

[0007] To achieve the above objectives, the present invention provides a laser cutting optimization method for mass production of flexible tactile sensors, comprising: obtaining flexible material characteristic data, configuring an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generating a cutting beam and a thermal annealing beam; generating plasma through a plasma generator, and guiding the plasma to a laser cutting area to improve laser energy absorption efficiency; collecting temperature field distribution, plasma density distribution, and cutting front shape data of the cutting process, and estimating cutting process state parameters based on the temperature field distribution, plasma density distribution, and cutting front shape data; dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma based on the cutting process state parameters; obtaining a cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy based on the cutting quality evaluation result; achieving simultaneous cutting of multiple sensor units according to the control strategy, and performing a post-cutting processing process to complete mass production of flexible tactile sensors.

[0008] Furthermore, the ultrafast laser cutting unit has at least two laser sources. When the ultrafast laser cutting unit is working, one of the low-power laser sources realizes one cutting, and the cutting depth reaches one-third to one-half, forming a relatively flat incision base; the remaining depth of cutting is completed by the remaining one or more laser sources.

[0009] Furthermore, the method of obtaining the flexible material characteristic data, configuring the ultrafast laser cutting unit and the thermal annealing laser unit according to the flexible material characteristic data, and generating the cutting beam and the thermal annealing beam includes: identifying the type of flexible material through spectral analysis and image recognition technology, obtaining the absorption, reflection and thermal conduction characteristic data of the flexible material to lasers of different wavelengths, and generating a flexible material characteristic database; matching the optimal laser parameter combination according to the flexible material characteristic database, configuring a femtosecond laser and a tunable wavelength continuous laser, and generating the cutting beam and the thermal annealing beam; and setting the timing, energy and spatial position of each beam through an optoelectronic control unit, and adjusting the optical reflector and the electrically controlled displacement platform to achieve coordination between the cutting beam and the thermal annealing beam.

[0010] Furthermore, the plasma is generated by a plasma generator and guided to the laser cutting area to improve the laser energy absorption efficiency, including: selecting a working gas according to the type of the flexible material, generating plasma in the cutting area through radio frequency discharge or microwave discharge technology; guiding the plasma through a plasma transmission channel, adjusting the discharge power and gas flow, and controlling the density parameters and temperature parameters of the plasma.

[0011] Furthermore, guiding the plasma to the laser cutting area to improve the laser energy absorption efficiency includes: receiving the plasma generated by the plasma generator, guiding the plasma through an electromagnetic field to ensure that it coincides with the laser cutting area; monitoring the interaction between the plasma and the laser in the laser cutting area to obtain energy transfer efficiency data; and optimizing the synergistic effect of the plasma and laser cutting based on the energy transfer efficiency data.

[0012] Furthermore, the temperature field distribution, plasma density distribution and cutting front shape data of the cutting process are collected, and the cutting process state parameters are estimated based on the temperature field distribution, plasma density distribution and cutting front shape data, including: collecting the temperature field, plasma density and cutting front shape data of the cutting process to form a multidimensional sensing data set; receiving the multidimensional sensing data set, performing data preprocessing and consistency verification, and generating an effective monitoring data stream; based on the effective monitoring data stream, establishing a state observation equation using discrete-time linear time-invariant system theory, constructing a distributed unknown input observer model, and estimating the initial cutting process state parameters based on the distributed unknown input observer model; decomposing the initial cutting process state parameters into multiple subsystems, and realizing the estimation of unknown input disturbances through parallel processing of a distributed computing architecture to obtain multi-source state estimation results; performing weighted fusion on the multi-source state estimation results, evaluating the reliability of the estimation results, and outputting the cutting process state parameters.

[0013] Furthermore, the method dynamically adjusts the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma, based on the cutting process state parameters, including: receiving the cutting process state parameters, comparing them with preset cutting quality standards, and identifying adjustment parameter items; establishing a plasma-laser-material interaction model based on the adjustment parameter items, predicting the optimal parameter combination through a machine learning algorithm, and forming parameter adjustment instructions; and controlling the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma according to the parameter adjustment instructions, thereby achieving real-time optimization of the cutting process.

[0014] Furthermore, the method of acquiring a cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy based on the cutting quality evaluation result includes: acquiring the cutting edge image, applying an edge detection algorithm to extract edge features, and obtaining edge quality data; receiving the edge quality data, comparing it with a preset standard quality template, and calculating a cutting quality score; receiving the cutting quality score, determining whether it reaches a preset quality score threshold, and forming a control decision request; querying a preset control strategy library based on the control decision request, selecting the most matching control strategy or reconstructing a new control strategy, and generating a cutting control scheme; evaluating the cutting control scheme, obtaining a cutting control scheme that has passed the evaluation, and storing the cutting control scheme that has passed the evaluation in a knowledge base for continuous optimization of the control strategy library.

[0015] Furthermore, the control strategy is followed to achieve simultaneous cutting of multiple sensor units, and post-cutting processing is performed to complete the mass production of flexible tactile sensors, including: configuring a multi-station parallel cutting system according to the control strategy to achieve simultaneous cutting of multiple sensor units; real-time monitoring of cutting quality, and establishing a cutting-detection-adjustment closed-loop process flow to ensure quality consistency in mass production.

[0016] Furthermore, the post-cutting processing process is carried out to complete the mass production of flexible tactile sensors, including: using plasma cleaning or micro-thermal treatment technology to optimize the physical and chemical properties of the cutting edges of the sensor units cut by the multi-station parallel cutting system; and realizing the full-process data collection, analysis and traceability of flexible materials through a full-process digital management system to complete the mass production of high-quality flexible tactile sensors.

[0017] The present invention also provides a laser cutting optimization system for mass production of flexible tactile sensors, comprising:

[0018] A multi-beam laser collaborative cutting unit is used to obtain flexible material characteristic data, configure an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generate a cutting beam and a thermal annealing beam;

[0019] A plasma assist unit, configured to generate plasma through a plasma generator and guide the plasma to a laser cutting area to improve laser energy absorption efficiency;

[0020] A distributed state observation unit is used to collect temperature field distribution, plasma density distribution and cutting front shape data during the cutting process, and estimate cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data;

[0021] a parameter collaborative optimization unit, configured to dynamically adjust the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, and the density and temperature of the plasma based on the cutting process state parameters;

[0022] a cutting quality assessment unit, configured to acquire a cutting edge image, assess the cutting quality, obtain a cutting quality assessment result, and select or reconstruct a control strategy based on the cutting quality assessment result;

[0023] The batch production unit is used to realize the simultaneous cutting of multiple sensor units according to the control strategy, and perform post-cutting processing to complete the batch production of flexible tactile sensors.

[0024] The beneficial effects of the present invention include:

[0025] 1. Through multi-beam laser collaborative cutting technology, high-energy-density ultrafast laser is combined with low-energy continuous laser to achieve simultaneous cutting and thermal annealing, effectively solving the edge damage problem in traditional single laser cutting;

[0026] 2. Plasma-assisted laser cutting technology is used to change the surface characteristics of the material by introducing specific plasma, thereby improving the laser energy absorption efficiency, reducing the required laser power, and reducing the heat-affected zone and material deformation;

[0027] 3. A distributed unknown input observer was designed based on discrete-time linear time-invariant system theory to achieve real-time monitoring and state estimation of key parameters of the cutting process, providing accurate feedback for intelligent control;

[0028] 4. A reconfigurable intelligent surface control system was constructed to dynamically adjust the cutting strategy based on observer feedback, achieving adaptive control for different materials and cutting requirements;

[0029] 5. Developed a multi-parameter collaborative optimization algorithm, established a plasma-laser-material interaction model, and optimized parameter combinations through machine learning methods to achieve a balance between cutting quality and efficiency;

[0030] 6. Realizes batch parallel cutting process. Through multi-station layout and coordinated control, it can realize simultaneous cutting of multiple sensor units, greatly improving production efficiency.

[0031] 7. Established full-process digital management technology to realize data collection and analysis of the entire process from material entry to finished product delivery, ensuring quality consistency and traceability of batch production. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0033] Figure 1 This is a structural block diagram of the laser cutting optimization system for mass production of flexible tactile sensors of the present invention;

[0034] Figure 2 A schematic flow chart of the laser cutting optimization method for mass production of flexible tactile sensors according to the present invention;

[0035] Figure 3 This is a schematic structural diagram of a multi-beam laser collaborative cutting unit of the present invention;

[0036] Figure 4 is a workflow diagram of the cutting quality assessment unit of the present invention;

[0037] Figure 5 Schematic diagram of the layout of the mass production unit of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0039] like Figure 1 As shown, the present invention provides a laser cutting optimization system for mass production of flexible tactile sensors, including: a multi-beam laser collaborative cutting unit 10, a plasma assist unit 20, a distributed state observation unit 30, a parameter collaborative optimization unit 40, a cutting quality evaluation unit 50 and a mass production unit 60.

[0040] The multi-beam laser collaborative cutting unit 10 is used to obtain flexible material characteristic data, configure the ultrafast laser cutting unit and the thermal annealing laser unit according to the flexible material characteristic data, and generate a cutting beam and a thermal annealing beam;

[0041] A plasma assist unit 20 is used to generate plasma through a plasma generator and guide the plasma to the laser cutting area to improve the laser energy absorption efficiency;

[0042] A distributed state observation unit 30 is used to collect temperature field distribution, plasma density distribution and cutting front shape data during the cutting process, and estimate cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data;

[0043] a parameter collaborative optimization unit 40 for dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma, based on the cutting process state parameters;

[0044] a cutting quality evaluation unit 50 for acquiring a cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy according to the cutting quality evaluation result;

[0045] The batch production unit 60 is used to realize the simultaneous cutting of multiple sensor units according to the control strategy, and perform post-cutting processing to complete the batch production of flexible tactile sensors.

[0046] In this embodiment, the ultrafast laser cutting unit features at least two laser sources. During operation, one low-power laser source performs a single cut, reaching one-third to one-half the cutting depth, creating a relatively flat cut base. The remaining laser source or sources complete the remaining cutting depth. This collaboration solves issues such as slow cutting speeds caused by a single low-power source and the appearance of a burnt, rough, or blackened cut surface caused by a single high-power source.

[0047] like Figure 2 As shown, in one embodiment, the laser cutting optimization method for mass production of flexible tactile sensors provided by the present invention includes the following steps:

[0048] Step S1: Acquire flexible material characteristic data, configure an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generate a cutting beam and a thermal annealing beam.

[0049] In this step, a spectrum analyzer is first used to scan the flexible material in real time to collect its spectral reflection characteristics, while a high-resolution image sensor is used to obtain the surface texture and structural characteristics of the material. The collected spectral data is compared with a pre-established material property database to quickly identify the type of material currently being processed. For the identified material, the absorption rate curve of the material for different wavelengths of lasers, the thermal conductivity coefficient, and key parameters such as the melting point and vaporization point of the material are automatically retrieved to form a targeted processing parameter set. Based on these parameters, the system automatically matches the optimal laser parameter combination from the laser parameter library, including parameters such as wavelength, pulse width, pulse energy, and repetition frequency.

[0050] Based on the identified characteristics of the flexible material, the operating parameters of the ultrafast laser cutting unit will be automatically configured. Specifically, the system will adjust the output power, pulse repetition frequency and pulse width of the femtosecond laser according to the thickness and composition of the material, and adjust the spatial energy distribution of the beam through the beam shaping system to ensure the uniformity of the energy density during the cutting process. At the same time, the parameters of the thermal annealing laser unit will be configured according to the thermophysical properties of the material, including the wavelength selection, output power and spot size of the continuous laser. The wavelength selection of the thermal annealing laser will be optimized according to the absorption spectrum characteristics of the material to ensure that the energy can be effectively absorbed by the material to form an appropriate heat treatment effect.

[0051] Precise control of each beam is achieved through a central optoelectronic control unit. This control unit sets the timing relationship between the beams, namely the delay between the ultrafast laser cutting pulse and the thermal annealing laser, based on the optimal parameters calculated by the optimization algorithm. It also adjusts the energy distribution of each beam to ensure that excessive energy does not cause excessive material damage during the cutting process. It also controls the relative spatial positions of the beams to achieve the optimal spatial configuration of the cutting and thermal annealing beams within the work area. This configuration typically involves the thermal annealing beam surrounding the cutting beam, forming a spatial sequence of preheating, cutting, and annealing. To achieve this precise spatial layout, the system uses high-precision optical mirrors and an electrically controlled displacement platform to adjust the position and angle of each beam at the micron level, ensuring that the cutting and thermal annealing beams work precisely together.

[0052] Specifically, the method of obtaining flexible material characteristic data, configuring an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generating a cutting beam and a thermal annealing beam includes: identifying the type of flexible material through spectral analysis and image recognition technology, obtaining absorption, reflection, and thermal conduction characteristic data of the flexible material to lasers of different wavelengths, and generating a flexible material characteristic database; matching an optimal laser parameter combination according to the flexible material characteristic database, configuring a femtosecond laser and a tunable wavelength continuous laser, and generating the cutting beam and the thermal annealing beam; and setting the timing, energy, and spatial position of each beam through an optoelectronic control unit, and adjusting an optical reflector and an electrically controlled displacement platform to achieve coordination between the cutting beam and the thermal annealing beam.

[0053] During the specific implementation of this step, the flexible material to be processed is first scanned in real time by a multifunctional spectral analysis device. This set of equipment includes a multi-band reflectance spectrometer, which covers a wide spectrum from ultraviolet to far infrared (190nm-2500nm), and can capture the characteristic reflection peaks of different flexible materials at various wavelengths. At the same time, the system is also equipped with a Raman spectrometer to obtain molecular vibration information of the material and accurately identify the polymer type and modifier. For composite flexible materials, the system will also use attenuated total reflection (ATR) infrared spectroscopy technology to analyze the chemical composition of the material surface to ensure accurate identification of multi-layer structure materials.

[0054] Simultaneously, a high-resolution image acquisition system captures the microscopic topography of the material surface from multiple angles. Equipped with a macro lens and an adjustable lighting array, this imaging system is capable of capturing the texture, color, and reflective properties of the material surface under varying lighting conditions. After preprocessing, the captured images are fed into a pretrained deep convolutional neural network for analysis. This network, trained on a large number of flexible material samples, can identify material type from image features with 98% accuracy. The system cross-validates the results of spectral analysis and image recognition to ensure accurate material type determination, particularly for materials with similar appearances but different compositions.

[0055] Once the material type is determined, the complete characteristic data set of the material will be immediately retrieved from the database. This data set contains the absorption coefficient curve of the material for lasers of different wavelengths (from ultraviolet to far infrared), reflectivity distribution, thermal conductivity coefficient, thermal diffusion coefficient, phase transition temperature and other key physical parameters. For new materials that have not yet been included in the database, the system will start the rapid characteristic measurement module, obtain the necessary parameters through sample testing, and add the results to the database to achieve continuous expansion of the knowledge base. The system will also consider the thickness, number of layers and structural characteristics of the material to construct a three-dimensional thermal conduction model of the material and predict the heat distribution under the action of the laser.

[0056] Based on the acquired material property data, the optimization algorithm begins to calculate the optimal laser parameter combination. This process uses a multi-objective optimization algorithm, taking into account the three goals of cutting quality, efficiency and heat-affected zone control. The system simulates the laser-material interaction process under different parameters and calculates the energy deposition distribution, heat diffusion and material removal efficiency. For the ultrafast laser cutting unit, the system will optimize the selection of the most suitable central wavelength (usually between 1030nm-1064nm), pulse width (typically 200-500 femtoseconds), pulse energy and repetition frequency. For the thermal annealing laser unit, the system will select the optimal wavelength from the optional wavelengths of the tunable wavelength continuous laser (usually covering 780nm-1550nm) according to the spectral absorption characteristics of the material, and calculate the required power density and irradiation time to achieve the best thermal annealing effect.

[0057] After the parameters are determined, the actual laser equipment is configured. For femtosecond lasers, the system sets the output power, pulse repetition rate, and compressor parameters to ensure that the output pulses reach the expected time width and energy. The laser output passes through a beam shaping system, which uses a combination of a spatial light modulator and a phase plate to adjust the original Gaussian distribution beam to a more uniform flat-top distribution or a specific custom energy distribution to optimize the cutting effect. The shaped beam passes through a high-precision focusing system, which uses a specially designed multi-element lens group to maintain a small focal point (up to 5 microns in diameter) while maintaining a large depth of focus to accommodate the tiny fluctuations that may exist in flexible materials.

[0058] For the thermal annealing laser unit, the appropriate wavelength is selected based on calculations, and precise wavelength output is achieved through a tunable diode laser or optical parametric oscillator. The thermal annealing beam is split into multiple paths by a beam splitter system, forming a heat treatment zone surrounding the main cutting beam. Each beam is equipped with an independent power regulator and focusing optics, enabling the system to create different thermal field distributions as needed, such as using a preheating mode at the cutting front and an annealing mode in the post-cut area.

[0059] The most complex part is achieving precise coordination of multiple laser beams. As the center of the system, the optoelectronic control unit is responsible for the unified scheduling of the timing, energy, and spatial position of each beam. In terms of timing control, the system uses a high-precision delay generator that can control the time relationship between the ultrafast laser pulse and the thermal annealing laser with picosecond precision, ensuring that the thermal annealing beam reaches the working area within the optimal time window after the cutting beam acts. In terms of energy control, the system uses a closed-loop feedback mechanism to monitor the actual energy of each beam in real time, and quickly adjusts it through an electro-optic modulator or an acousto-optic modulator to maintain the stability and precise proportion of the energy of each beam.

[0060] For spatial position control, the system utilizes a combination of high-precision optical mirrors and electrically controlled displacement stages. Driven by piezoelectric ceramics or acousto-optic deflectors, these mirrors offer fast response times and high positioning accuracy, enabling microsecond-level adjustments of the beam position. The electrically controlled displacement stages, powered by linear motors, handle larger-scale position adjustments, achieving nanometer-level positioning accuracy. Through the coordinated operation of these precision mechanisms, the system enables precise positioning and relative adjustment of the cutting and thermal annealing beams in three dimensions, ensuring they maintain optimal spatial alignment, whether performing static or high-speed dynamic cutting.

[0061] The entire configuration process, from material identification to parameter optimization and equipment configuration, is automated by a central control system, eliminating the need for human intervention. This significantly reduces setup time and improves production efficiency. The system also records the parameters and results of each configuration in a material database and continuously optimizes the parameter prediction model using machine learning algorithms, improving the accuracy and efficiency of future configurations.

[0062] like Figure 3 As shown, the multi-beam laser collaborative cutting unit 10 includes: a femtosecond laser 11 for generating ultrafast laser pulses; a beam shaping system 12 for adjusting the spatial distribution of the laser beam; a high-precision focusing system 13 for focusing the laser beam onto the working area; an adjustable wavelength continuous laser 14 for generating a thermal annealing beam; a spectroscopic system 15 for dividing the thermal annealing beam into multiple paths; an optical path control system 16 for adjusting the relative position and angle of each beam; an optoelectronic control unit 17 for controlling the timing and energy of each beam; an optical reflector 18 and an electrically controlled displacement platform 19 for precisely controlling the spatial position of the beam.

[0063] Step S2: Generate plasma through a plasma generator, and guide the plasma to the laser cutting area to improve the laser energy absorption efficiency.

[0064] In this step, the most appropriate working gas combination is selected from a gas database based on the flexible material type identified in the previous step. For different flexible materials, the system selects different gas compositions and ratios to create the plasma environment most suitable for processing that material. For example, for oxygen-containing organic materials, a mixture of argon and a small amount of oxygen might be selected; for oxygen-sensitive materials, pure argon or an argon-nitrogen mixture might be selected. The system uses precision gas flow controllers to regulate the flow of each component gas, ensuring precise control of the gas mixture ratio.

[0065] After selecting the appropriate working gas, the plasma generator is activated. Depending on the material properties and processing requirements, the system may choose to generate plasma using either radio frequency (RF) or microwave discharge technology. RF discharge typically operates at an industrial frequency of 13.56 MHz and is suitable for producing large, uniform, low-temperature plasmas. Microwave discharge, on the other hand, typically operates at a frequency of 2.45 GHz and can produce plasmas with higher energy densities. By adjusting the discharge power and gas flow rate, the density and temperature parameters of the generated plasma are controlled to ensure that it alters the material's surface properties and increases laser absorption without causing excessive thermal damage to the material.

[0066] The generated plasma needs to be precisely guided to the laser cutting area. A designed plasma transmission channel guides the plasma from the generator to the working area. This transmission process is achieved through a carefully designed electrode system and magnetic field confinement to ensure the concentration and stability of the plasma. Once in the working area, the system uses a precise electromagnetic field adjustment device to fine-tune the distribution of the plasma so that it precisely coincides with the laser cutting area. In addition, the system monitors the interaction between the plasma and the laser in real time, collects the emission spectrum of the working area through a spectral analysis system, analyzes the concentration and energy distribution of active particles in the plasma, and evaluates the energy transfer efficiency.

[0067] Based on the monitored energy transfer efficiency data, the synergistic effect of plasma and laser cutting is optimized in real time. This optimization includes adjusting the plasma density distribution, changing the relative relationship between the spatial position of the plasma and the laser cutting area, and adjusting the plasma composition and energy to ensure that the plasma can most effectively change the surface properties of the material, improve the efficiency of laser energy absorption, reduce the heat-affected zone, and improve cutting quality.

[0068] Specifically, the plasma is generated by a plasma generator and guided to the laser cutting area to improve the laser energy absorption efficiency, including: selecting a working gas according to the type of the flexible material, generating plasma in the cutting area through radio frequency discharge or microwave discharge technology; guiding the plasma through a plasma transmission channel, adjusting the discharge power and gas flow, and controlling the density parameters and temperature parameters of the plasma.

[0069] Furthermore, guiding the plasma to the laser cutting area to improve the laser energy absorption efficiency includes: receiving the plasma generated by the plasma generator, guiding the plasma through an electromagnetic field to ensure that it coincides with the laser cutting area; monitoring the interaction between the plasma and the laser in the laser cutting area to obtain energy transfer efficiency data; and optimizing the synergistic effect of the plasma and laser cutting based on the energy transfer efficiency data.

[0070] During the system's actual implementation, the plasma-assisted unit first receives flexible material type information from the preceding unit via a material data interface. Based on a pre-established gas-material matching database, the system selects the optimal working gas combination for each flexible material type. For oxygen-containing polymers such as PET and PC, the system typically uses an argon-oxygen mixture with an oxygen ratio controlled between 5% and 15%. This combination maintains a stable plasma state while enhancing the oxidative decomposition of the material through reactive oxygen species, improving cutting efficiency. For oxygen-sensitive materials such as PI and PVDF, the system uses pure argon or an argon-nitrogen mixture, typically with a nitrogen ratio of 10% to 30%, to avoid edge degradation caused by oxidation reactions. For fluoropolymers such as PTFE, the system uses a specially formulated helium-argon mixture, leveraging helium's high thermal conductivity and inertness to minimize the environmental impact of fluoride release. The system utilizes a high-precision gas mixing control unit to precisely mix up to four gases, with a mixing accuracy of ±0.5%, ensuring stable and consistent gas composition.

[0071] After gas selection is complete, the most appropriate plasma generation technology is determined based on the material thickness and thermal conductivity. For thin materials (thickness <0.3mm) or heat-sensitive materials, the system preferentially uses radio frequency (RF) discharge technology, typically operating at a frequency of 13.56MHz. This technology produces a relatively low temperature (2000-5000K) uniform plasma, avoiding excessive thermal effects. For medium-thick materials (0.3-2mm), the system uses microwave discharge technology, typically operating at a frequency of 2.45GHz, which can produce a higher energy density plasma (5000-8000K) and provide a stronger material activation effect. For particularly thick or difficult-to-cut composite materials, the system uses RF-microwave hybrid discharge technology, combining the advantages of both methods to provide sufficient energy density while maintaining stability. The discharge chamber adopts a specially designed cavity structure, with an inner wall made of high-temperature ceramic material and equipped with a water cooling system to ensure long-term stability. The cavity geometry has been carefully optimized to form a directional airflow channel to guide the plasma flow to the cutting area. Inside the cavity, the system arranges a multi-point electrode array. By controlling the voltage and phase of each electrode, it achieves precise control of the spatial distribution of the plasma and produces the plasma form that best suits the current cutting task.

[0072] The generated plasma needs to be guided to the cutting area through a specially designed transmission channel. This transmission channel is a multi-segment guide structure with an internal temperature gradient design. The temperature gradually decreases along the length of the channel, reducing the energy loss and composition changes of the plasma during transmission. There are multiple adjustment nodes inside the channel, each node is equipped with an electromagnetic coil and a pneumatic control valve. By adjusting the electromagnetic field intensity and the gas flow mode, the flow trajectory and diffusion range of the plasma can be precisely controlled. The outlet end of the channel is designed as a deformable nozzle, which can dynamically adjust the outlet shape according to changes in the cutting contour to ensure that the plasma can always accurately cover the cutting front area. In particular, for high-precision cutting tasks, the system will install a focusing device at the outlet of the channel to compress the plasma into a highly directional beam with a diameter as small as 0.5mm, which is suitable for precise cutting of fine structures.

[0073] Precise control of plasma density and temperature parameters is achieved through precise discharge power control and gas flow regulation. Discharge power control utilizes fully digital power modulation technology, offering an output power range of 50-2000W, with regulation accuracy better than ±5W and a response time of less than 10ms, enabling rapid adjustment of discharge intensity based on real-time feedback. The gas flow control system utilizes an array of thermal mass flow controllers, independently controlling each gas flow rate from 0.1 to 20L / min with control accuracy better than ±1%, enabling precise regulation of total gas flow rate and mixing ratio. By synergistically controlling these two key parameters, the system precisely adjusts the plasma density distribution (typically in the range of 10^15-10^17 cm^-3) and temperature gradient (core temperature 2000-10000K). Specifically, increasing discharge power while reducing gas flow produces a high-temperature, high-density plasma suitable for thick or difficult-to-cut materials; whereas reducing power and increasing flow produces a lower-temperature plasma with a larger coverage area, suitable for processing heat-sensitive materials. The system will dynamically adjust these parameters based on real-time material temperature feedback and cutting progress to ensure cutting efficiency while avoiding overheating damage.

[0074] After the plasma flows out of the transmission channel, the system's electromagnetic field guidance device begins to take over its precise control. This device includes multiple sets of orthogonally arranged electromagnetic coils and high-frequency electric field generators. By generating a precisely controlled electromagnetic field distribution, it exerts a directional force on the charged plasma particles and adjusts their motion trajectory. The system adopts a real-time closed-loop control strategy. Based on the plasma position and morphology information collected by the high-speed camera and spectrometer, it calculates the deviation between the current plasma distribution and the ideal state, and then adjusts the current size and direction of each electromagnetic coil to correct the spatial distribution of the plasma in real time to ensure that it accurately coincides with the laser cutting area. In specific implementation, the system uses multi-level electromagnetic field control, including far-field guidance, mid-field focusing and near-field fine adjustment. Three levels realize the full process control of the plasma from coarse adjustment to fine adjustment. Even during high-speed cutting (cutting speed can reach 500mm / s), the system can still maintain dynamic synchronization between the plasma and the cutting front, with a tracking accuracy better than 0.1mm, ensuring that the plasma always acts on the area where it is most needed.

[0075] To ensure optimal synergy between plasma and laser, a dedicated interaction monitoring module was designed. This module comprises a high-time-resolution spectral analysis system, a plasma diagnostic probe array, and a material temperature monitoring system. The spectral analysis system utilizes a fiber array acquisition system and a grating spectrometer to acquire the emission spectrum of the cutting area with microsecond time resolution, from which key parameters such as plasma temperature, electron density, and excited-state distribution are extracted. Specifically, the system monitors specific spectral line intensity ratios, such as the ratio of the argon 4p-4s transition line to the background continuum spectrum. This ratio directly reflects the efficiency of laser energy absorption and conversion by the plasma. A plasma diagnostic probe array is positioned around the cutting area to collect plasma data such as current, potential distribution, and ion flux. These data reflect the plasma activity and the intensity of its interaction with the material. The material temperature monitoring system uses high-speed infrared thermal imaging technology to capture the temperature field distribution and its temporal evolution in the cutting area, directly reflecting the energy deposition and conduction in the material.

[0076] These monitoring data are processed and fused in real time to form a comprehensive energy transfer efficiency data set. The system calculates multiple key indicators, including the conversion efficiency of laser energy to plasma, the transfer efficiency of plasma energy to material, the cutting efficiency of the material (material removal rate per unit energy input), and the spatial distribution efficiency of energy utilization. These indicators together constitute a complete description of the plasma-laser-material energy transfer chain, revealing the efficiency bottleneck of energy flow and conversion under the current configuration. Based on these efficiency data, the system combines preset optimization goals (such as maximizing cutting speed, minimizing heat-affected zone, or a balance between the two) to determine the optimization direction and generate specific adjustment strategies.

[0077] Based on the energy transfer efficiency data, it enters the synergistic effect optimization stage. This process uses an adaptive optimization algorithm to comprehensively consider the complex interactions between multiple control parameters. First, the system will adjust the synergistic relationship between plasma and laser in the time domain, including plasma pretreatment time (before laser action), plasma-laser coexistence time and plasma post-processing time (after laser action), and find the optimal timing configuration so that the plasma can provide the most effective auxiliary effect at the critical moment of laser cutting. Secondly, the system will optimize the synergistic relationship in the spatial domain and adjust the direction, size and shape of the plasma flow to maximize the laser cutting effect, while avoiding excessive diffusion of plasma into non-cutting areas to cause unnecessary thermal effects. Thirdly, the system will optimize the synergistic relationship in the energy domain, and adjust the distribution ratio between laser energy and plasma energy while keeping the total energy input unchanged, to find the combination with the highest energy utilization efficiency.

[0078] Through this comprehensive optimization and adjustment, the best synergy between plasma and laser cutting can be achieved. Under the optimized configuration, plasma can not only significantly improve the absorption efficiency of laser energy (the improvement can reach 30% to 50%), but also change the surface chemical state of the material, reduce the decomposition energy consumption, and at the same time blow away the debris and melt generated by cutting through the airflow effect to prevent edge contamination caused by redeposition. This synergistic effect enables the system to maintain or improve cutting efficiency while reducing laser power, greatly reducing the scope of the heat-affected zone (reduction of up to 40% to 60%) and material deformation, thereby significantly improving the cutting quality and production stability of the flexible tactile sensor. The system will continuously monitor and optimize this synergistic effect to ensure that it always maintains the best working state when material properties change or process conditions fluctuate.

[0079] The plasma assist unit 20 includes: a plasma generator for generating low-temperature plasma; a gas supply system for providing the required working gas; a plasma transmission channel for guiding the plasma to the cutting area; an electromagnetic field adjustment device for controlling the spatial distribution of the plasma; a density and temperature control unit for adjusting the physical parameters of the plasma; and a spectral analysis system for monitoring the composition and state of the plasma.

[0080] Step S3: collecting temperature field distribution, plasma density distribution and cutting front shape data of the cutting process, and estimating cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data.

[0081] During this step, a variety of sensors are deployed to comprehensively monitor the key parameters of the cutting process. A high-speed infrared thermal imager scans the cutting area in real time to acquire temperature field distribution data. The temperature resolution can reach 0.1°C and the spatial resolution can reach 50 microns, accurately capturing the instantaneous temperature changes and spatial distribution during the cutting process. Simultaneously, an optical emission spectrometer monitors the intensity and distribution of the plasma emission spectrum, thereby inferring the plasma density distribution. Furthermore, the system uses high-speed cameras and structured light scanning technology to capture the shape and evolution of the cutting front in real time and analyze the cutting quality and speed. The data collected by these sensors is synchronously acquired through a high-speed data acquisition system, forming a multidimensional sensor data set for the cutting process.

[0082] The acquired multidimensional sensor data first passes through the data preprocessing module. This module uses digital filtering techniques to remove signal noise and perform data calibration and calibration to ensure temporal synchronization and spatial correspondence between different sensor data. This data is then subjected to consistency checks to identify and remove anomalous data points, ensuring data validity and reliability. This processed data forms an effective monitoring data stream for subsequent state estimation.

[0083] Based on the effective monitoring data stream, the state observation equations of the cutting process are established using discrete-time linear time-invariant system theory. These equations describe the relationship between key state parameters of the cutting process (such as cutting depth, cutting speed, heat-affected zone width, etc.) and observable signals (temperature field, plasma density, cutting front shape, etc.). Based on these equations, the system constructs a distributed unknown input observer model, which can accurately estimate the system state in the presence of unknown interference and measurement noise. Using this model, the system estimates the initial state parameters of the cutting process, including the actual cutting depth, material removal rate, heat-affected zone range, etc.

[0084] To improve computational efficiency and robustness, the initial state parameters are decomposed into multiple subsystems, such as the cutting dynamics subsystem, the heat transfer subsystem, and the plasma impact subsystem. These subsystems are processed in parallel using a distributed computing architecture, with each subsystem responsible for estimating a specific set of state parameters. This parallel processing enables the system to complete state estimation within milliseconds, meeting the requirements of real-time control. This distributed processing approach also improves the system's fault tolerance, enabling it to provide reliable state estimates even when some sensor data is missing or individual observers fail.

[0085] Finally, the multi-source state estimation results generated by each subsystem are weighted and fused. This fusion process considers the reliability and uncertainty of each estimation result, using an adaptive weighting method to assign higher weights to more reliable estimates. The overall reliability of the fusion results is also evaluated, and the uncertainty level of the state estimation is quantified by calculating the estimation error covariance matrix. Ultimately, the system outputs high-confidence cutting process state parameters, providing accurate feedback information for subsequent parameter optimization and control.

[0086] Specifically, the temperature field distribution, plasma density distribution and cutting front shape data of the cutting process are collected, and the cutting process state parameters are estimated according to the temperature field distribution, plasma density distribution and cutting front shape data, including: collecting the temperature field, plasma density and cutting front shape data of the cutting process to form a multidimensional sensing data set; receiving the multidimensional sensing data set, performing data preprocessing and consistency verification, and generating an effective monitoring data stream; based on the effective monitoring data stream, establishing a state observation equation using discrete-time linear time-invariant system theory, constructing a distributed unknown input observer model, and estimating the initial cutting process state parameters based on the distributed unknown input observer model; decomposing the initial cutting process state parameters into multiple subsystems, and realizing the estimation of unknown input disturbances through parallel processing of a distributed computing architecture to obtain multi-source state estimation results; performing weighted fusion on the multi-source state estimation results, evaluating the reliability of the estimation results, and outputting the cutting process state parameters.

[0087] Among them, the distributed state observation unit 30 includes: a multi-sensor acquisition module for collecting temperature field, plasma density and cutting front shape data during the cutting process; a data preprocessing module for data filtering and calibration; a state parameter model construction module for establishing the state observation equation; a distributed computing module for decomposing the state parameters into multiple subsystems for processing; a state estimation module for performing unknown input disturbance estimation; and a result evaluation and fusion module for weighted fusion of multi-source state estimation results.

[0088] Step S4: dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, and the density and temperature of the plasma based on the cutting process state parameters.

[0089] In this step, the cutting process state parameters are first received from the distributed state observation unit. These parameters include key indicators such as cutting depth, cutting speed, heat-affected zone width, and material removal rate. The system compares and analyzes these real-time state parameters with the preset cutting quality standards. These quality standards are pre-defined based on the functional requirements and performance indicators of different flexible tactile sensors, including cutting edge roughness, maximum allowable width of the heat-affected zone, cutting accuracy deviation range, etc. Through comparative analysis, the system identifies the parameters that need to be adjusted. For example, when the heat-affected zone exceeds the preset range, the system will identify the need to adjust the laser energy or thermal annealing beam parameters; when the cutting speed is insufficient, the plasma density or laser pulse frequency may need to be adjusted.

[0090] Based on the identified adjustment parameter items, the system activates the parameter optimization module. This module first establishes a plasma-laser-material interaction model to describe the complex energy transfer and material transformation relationship between the three. This model comprehensively considers physical phenomena such as the absorption and conduction of laser energy, the ionization and energy transfer effects of plasma, and the phase change and gasification process of the material. The system uses machine learning algorithms, especially methods based on deep reinforcement learning, to predict the cutting effect under different parameter combinations based on this model. Through training with a large amount of historical cutting data, it can quickly predict which parameter combination will produce the best cutting effect under the current material and state. Based on these prediction results, the system forms specific parameter adjustment instructions.

[0091] Based on the generated parameter adjustment instructions, the control unit begins to perform the actual parameter adjustments. For the cutting beam, the system may adjust the laser's output power, pulse repetition frequency, and pulse width; for the thermal annealing beam, the system may adjust its energy density, spot size, and relative position; for plasma, the system may adjust the discharge power, gas flow rate, and component ratio to change the plasma density and temperature. These adjustments are implemented through a precise electronic control system with a response time of milliseconds, ensuring that it can adapt to rapidly changing cutting conditions.

[0092] Specifically, the system optimizes the synergistic effects between various parameters. For example, when increasing the cutting beam energy to improve cutting efficiency, the system simultaneously adjusts the energy and position of the thermal annealing beam to compensate for the potential increase in the heat-affected zone. When changing the plasma density, the laser parameters are adjusted accordingly to maintain optimal energy absorption efficiency. This multi-parameter collaborative optimization ensures that the system improves cutting efficiency without sacrificing cutting quality.

[0093] The system also records the effects of each parameter adjustment, including state changes and improvements in cutting quality before and after the adjustment. These records are stored in a parameter optimization knowledge base, used to continuously refine the parameter optimization model and improve the accuracy of future predictions. Through this closed-loop learning mechanism, the system's parameter optimization capabilities will continue to improve with usage, enabling it to cope with increasingly complex cutting scenarios and material changes.

[0094] Specifically, the method of dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma based on the cutting process state parameters includes: receiving the cutting process state parameters, comparing them with preset cutting quality standards, and identifying adjustment parameter items; establishing a plasma-laser-material interaction model based on the adjustment parameter items, predicting the optimal parameter combination through a machine learning algorithm, and forming parameter adjustment instructions; and controlling the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma according to the parameter adjustment instructions to achieve real-time optimization of the cutting process.

[0095] When establishing the plasma-laser-material interaction model, a multi-physics field coupling equation system is used to represent the relationship between the three. This model includes the laser energy transfer equation, the plasma physical parameter evolution equation, and the material thermodynamic response equation. The mathematical expression is as follows:

[0096] The laser energy transfer process is described by the following formula:

[0097] E(r,t)=E0·exp(-αd)·f(r)·g(t)

[0098] Where E(r,t) represents the energy distribution at spatial position r and time t, E0 is the initial laser energy, α is the material absorption coefficient, d is the propagation distance, f(r) and g(t) are the spatial and temporal distribution functions, respectively.

[0099] The plasma density evolution equation is:

[0100]

[0101] Where n is the plasma density, S is the source term (related to the laser energy), αr is the recombination coefficient, and v is the plasma velocity.

[0102] The material temperature field distribution adopts the heat conduction equation:

[0103]

[0104] Where ρ is the material density, c is the specific heat capacity, T is the temperature, κ is the thermal conductivity, and Q is the heat source term (related to the laser energy and plasma action).

[0105] Based on the above model, a deep reinforcement learning algorithm was used for parameter optimization. Specifically, a Double Deep Q-Network (DDQN) architecture was used, consisting of two identical neural networks: a primary network and a target network. Each network consists of four layers: an input layer (32 nodes), two hidden layers (64 and 32 nodes, respectively), and an output layer. The activation function is ReLU.

[0106] The state space is defined as a 17-dimensional vector, which includes the temperature values ​​of key points of the temperature field (5 points), the characteristic values ​​of the plasma density distribution (5 points), the shape parameters of the cutting front (5 characteristic points), the current cutting speed, and the current width of the heat-affected zone.

[0107] The action space is defined as a discretized combination of parameter adjustments, including cutting beam energy (five levels), thermal annealing beam energy (five levels), the relative position of the two beams (three configurations), and plasma density (three levels), for a total of 225 combinations.

[0108] The reward function is designed as:

[0109] Among them, Q cut Score the cutting quality (0-10), W HAZ is the width of the heat affected zone, T process is the processing time, S edge is the edge smoothness, w1 to w4 are weight coefficients.

[0110] Training data is obtained through two channels: first, high-quality samples are screened from historical cutting records, totaling approximately 1,000 parameter-result pairs; second, online learning is performed, recording the state-action-reward sequence during each actual cutting process. An experience replay mechanism is employed, maintaining a replay buffer with a capacity of 10,000, and randomly sampling 256 batches for parameter updates during each training session.

[0111] Training was performed using the Adam optimizer, with an initial learning rate of 0.001 and a learning rate decay strategy. To balance exploration and exploitation, an ε-greedy strategy was employed, with an initial ε value of 0.9 and a linear decay to 0.05 over training. The model was trained for at least 50,000 steps, and convergence was determined when the average reward increase over 5,000 consecutive steps was less than 1%.

[0112] The parameter collaborative optimization unit 40 includes: a parameter comparison module 41, which is used to compare state parameters with quality standards; an interaction model construction module 42, which is used to establish a plasma-laser-material interaction model; a machine learning optimization module 43, which is used to predict the optimal parameter combination; a parameter control execution module 44, which is used to execute parameter adjustment instructions; and a parameter history recording module 45, which is used to store parameter optimization records to support system learning.

[0113] Step S5: Acquire a cutting edge image, evaluate the cutting quality, obtain a cutting quality evaluation result, and select or reconstruct a control strategy based on the cutting quality evaluation result.

[0114] In this step, high-resolution industrial cameras capture edge images of the completed cut area. These cameras are equipped with specialized optical systems that can capture detailed features of the cut edge from multiple angles and under varying lighting conditions. The captured raw images undergo preliminary processing by an image processing system, including image enhancement, noise removal, and contrast adjustment, to highlight the features of the cut edge. Advanced edge detection algorithms, such as the Canny algorithm or a deep learning-based edge detection network, are applied to extract precise edge contours from the processed images. These algorithms are capable of identifying edge positions at the sub-pixel level, ensuring detection accuracy down to the micron level. After edge detection, the edge's geometric characteristics, such as straightness, smoothness, and microscopic topography, are further analyzed to comprehensively evaluate the quality of the cut edge.

[0115] The extracted edge quality data is compared against pre-set standard quality templates. These templates are based on a collection of optimal edge features achieved under ideal cutting conditions and represent the highest quality level achievable for a specific material and cutting parameters. This comparison utilizes a multi-metric comprehensive scoring mechanism to quantify the gap between the current cut quality and the standard across multiple dimensions, including edge smoothness, perpendicularity, heat-affected zone width, and microcrack density. The system then weights and synthesizes the scores from these dimensions to calculate a comprehensive cut quality score that reflects the overall level of the current cut quality.

[0116] The system receives a cut quality score and compares it to a preset quality score threshold. This threshold represents the minimum quality standard required to meet product functional requirements. If the score falls below the threshold, the system marks the current cut quality as unsatisfactory and requires adjustment to the control strategy. If the score exceeds the threshold but is still far from the optimal score, the system marks the current cut as acceptable but with room for improvement. If the score is close to the optimal score, the system confirms that the current control strategy is effective. Based on this judgment, a specific control decision request is made, clarifying whether a minor adjustment to the current strategy is needed or a new control strategy needs to be selected or reconstructed.

[0117] Based on the control decision request, a preset control strategy library is queried. This strategy library contains a collection of control strategies for different materials, different cutting conditions, and different quality requirements. Each strategy consists of a series of specific operating parameters and control rules. A similar case retrieval algorithm is used to find the control strategy that best matches the current situation from the strategy library. If there is no fully matching strategy or the current quality score is particularly low, the strategy reconstruction module is activated to create a new control strategy by combining effective elements from existing strategies or introducing new control concepts. This reconstruction process is based on machine learning algorithms, which can extract successful experiences from historical data and avoid repeating previous mistakes.

[0118] The generated cutting control plan undergoes an internal system evaluation. Through simulation, the system predicts the plan's likely effectiveness under current conditions and calculates the risks and expected benefits of implementing the plan. If the evaluation results indicate that the plan can effectively improve cutting quality and that the risks are manageable, the system marks it as a successful cutting control plan and prepares it for implementation in the next cutting cycle. The plan and its evaluation results are also stored in the control knowledge base for future decision-making. This continuous learning and updating mechanism ensures that the control strategy library can be continuously optimized to adapt to new material properties and cutting requirements.

[0119] Specifically, the method of acquiring a cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy based on the cutting quality evaluation result includes: acquiring the cutting edge image, applying an edge detection algorithm to extract edge features, and obtaining edge quality data; receiving the edge quality data, comparing it with a preset standard quality template, and calculating a cutting quality score; receiving the cutting quality score, determining whether it reaches a preset quality score threshold, and forming a control decision request; querying a preset control strategy library based on the control decision request, selecting the most matching control strategy or reconstructing a new control strategy, and generating a cutting control scheme; evaluating the cutting control scheme, obtaining a cutting control scheme that passes the evaluation, and storing the cutting control scheme that passes the evaluation in a knowledge base for continuous optimization of the control strategy library.

[0120] During the cut quality assessment process, high-precision image acquisition equipment first captures microscopic images of the cut edge. This system comprises multiple collaborative imaging systems, including a high-resolution industrial camera, a microscope objective, and a specialized lighting system. The camera utilizes a global shutter CMOS sensor with a pixel size less than 2μm and a resolution exceeding 20 megapixels, ensuring the capture of micron-level edge details. The microscope objective offers variable magnification from 5x to 50x, adapting to inspection scenarios with varying precision requirements. The lighting system utilizes a multi-angle, multi-wavelength combination illumination solution, including a ring LED light source, a structured light projector, and a laser confocal light source. This allows the system to highlight various defects along the cut edge, such as microcracks, melt residue, and heat-affected discoloration, at various angles and under various conditions. Furthermore, the system is equipped with a 3D profile scanner, employing white light interferometry or laser triangulation to acquire 3D topographic data of the cut edge with a resolution of up to 50nm, providing precise data for edge verticality and roughness analysis.

[0121] The acquired raw image data first undergoes a series of preprocessing steps. An adaptive histogram equalization algorithm is applied to improve image contrast and enhance edge features. Non-local mean filtering or wavelet transform denoising methods are then used to effectively suppress image noise while preserving edge detail. Geometric correction is then performed to eliminate distortion caused by lens distortion and perspective differences. Finally, multiple images are registered and fused, integrating image information collected under different lighting conditions and angles into an enhanced composite image, providing comprehensive visual information for subsequent analysis. For 3D contour data, the system applies point cloud filtering and surface reconstruction algorithms to generate a highly accurate 3D edge model.

[0122] The preprocessed image data is fed into the edge feature extraction module. This module first applies a modified Canny edge detection algorithm, adaptively selecting the optimal threshold parameters to accurately identify the contour of the cut edge. The system then extracts various quantitative feature parameters along the identified edge contour: Fourier analysis and wavelet transform are used to calculate the edge's roughness index and spectral characteristics; morphological analysis is used to measure the edge's linearity deviation and local fluctuations; texture analysis is used to identify the density and distribution of microcracks and stress marks; color analysis is used to assess the extent and extent of the heat-affected zone; and 3D data analysis is used to calculate the edge's verticality, inclination angle, and surface roughness (Ra) value. The system also detects specific edge defects, such as melt beads, spatter, and incompletely cut areas, and records their location, size, and density. These extracted feature parameters collectively constitute a multidimensional data description of edge quality, comprehensively reflecting all aspects of cutting quality.

[0123] After receiving these edge quality data, they are scientifically compared with preset standard quality templates. These standard templates are pre-established quality reference standards for different types of flexible materials, different thicknesses, and different application scenarios, representing the best cutting quality achievable under ideal conditions. The comparison process adopts a multi-level weighted scoring mechanism: first, the system normalizes the deviation of each characteristic parameter from the corresponding standard value into a score in the range of 0-1; then, based on the different sensitivities of different application scenarios to quality characteristics, the system assigns different weight coefficients to each characteristic parameter. For example, for tactile sensor electrodes that require high conductivity, edge smoothness and heat-affected zone control will receive higher weights; while for structural components that require high flexibility, microcrack control and stress distribution will receive higher weights. The system calculates the comprehensive score of the cutting quality through a weighted summation method, and generates a quality analysis report in the form of a radar chart, which intuitively displays the scores of each indicator to help operators understand the advantages and disadvantages of the current cutting quality.

[0124] Once the quality score is calculated, the system immediately compares it against pre-set quality score thresholds. These thresholds are tiered to correspond to varying product quality requirements, ranging from the most basic functional readiness level to the most stringent high-precision application level. The judgment process not only considers the overall score but also checks whether key characteristic parameters meet minimum requirements. For example, even a high overall score will result in a failure if penetrating microcracks are present. Based on the comparison results, the system generates specific control decision requests: If the overall score falls below the minimum threshold, the system generates an "urgent adjustment" request, requesting an immediate halt to the current cutting process and significant parameter adjustments. If the score is between the minimum threshold and the desired threshold, the system generates an "optimization adjustment" request, recommending minor parameter adjustments while maintaining production continuity. If the score reaches or exceeds the desired threshold, the system generates a "maintain current" request, confirming that the current cutting parameters are valid. Furthermore, the system analyzes current quality trends based on historical data. If a gradual decline in quality is detected, even if it has not yet fallen below the threshold, an early warning signal will be issued, recommending preventative adjustments.

[0125] In response to a control decision request, a pre-defined control strategy library is consulted. This library, one of the system's core knowledge bases, contains a collection of strategies for addressing various materials, cutting issues, and quality deviations. These strategies are empirical rules derived from extensive experiments and production data analysis. Each strategy includes clear applicable conditions, parameter adjustment options, and expected results. The query process utilizes a hybrid approach of case-based reasoning (CBR) and fuzzy matching. The system first selects a set of candidate strategies from the strategy library based on the current material type, cutting conditions, and quality issue characteristics. The system then calculates the degree of match between each candidate strategy and the current situation, taking into account factors such as material similarity, quality issue similarity, and process condition similarity. Finally, the strategy with the highest match is selected as the baseline. If the highest match falls below a preset threshold, indicating that there are insufficient matching existing strategies in the strategy library, the system initiates a strategy reconstruction module. Using genetic algorithms or reinforcement learning methods, the system extracts effective elements from multiple related strategies and combines them into a new control strategy. This strategy reconstruction capability enables the system to handle unprecedented material combinations or cutting issues, demonstrating strong adaptability.

[0126] The selected or reconstructed control strategy is further refined and specified to form an executable cutting control plan. This plan encompasses several aspects: laser parameter adjustment, such as the specific adjustment values ​​for cutting and thermal annealing laser energy, frequency, and focus position; plasma parameter adjustment, such as discharge power, gas composition, and flow rate; cutting path and speed adjustment, optimizing the cutting sequence and speed profile; and special processing solutions, such as the application of pre-cutting and multiple scans in specific areas. The system generates detailed execution steps and a time sequence for the plan, ensuring a smooth transition and avoiding the instability that can result from significant parameter changes.

[0127] The generated cutting control plan isn't put into use directly; instead, it first enters an evaluation phase. The system first uses an internal simulation model to predict the potential effects of the plan's execution. This simulation model, based on a combination of physical principles and machine learning, simulates the laser-material interaction process and cutting results under different parameters. Through simulation, the system estimates the plan's expected quality improvement and potential risks, such as the risk of thermal damage from excessive energy input. The system also assesses the plan's operational feasibility, resource requirements, and impact on production efficiency. If the evaluation indicates that the plan can effectively improve current quality issues, that risks are manageable, and that there will be no significant impact on production efficiency, the system will mark it as an "evaluated" cutting control plan, preparing it for implementation in the next production cycle.

[0128] Finally, the approved cutting control schemes and their underlying decision logic are stored in the knowledge base. This storage process not only records the specific parameters of the scheme, but also includes the characteristics of the quality issues that triggered the scheme, the theoretical basis for the scheme, the expected results, and the feedback after actual implementation. The system regularly analyzes this historical record to evaluate the success rate and effectiveness stability of different strategies under different conditions, providing a basis for strategy rating and optimization. Through this continuous learning and knowledge accumulation mechanism, the control strategy library will continue to expand and optimize, and the applicability and effectiveness of the strategies will continue to increase with the system's use time. This self-evolutionary capability enables the system to adapt to changing material types and production requirements, maintaining long-term technological leadership.

[0129] During edge feature extraction, we used the multi-scale Canny edge detection algorithm combined with the deep learning edge segmentation network UNet++. The Canny algorithm parameters were set to a high threshold of 0.3, a low threshold of 0.1, and a Gaussian smoothing kernel size of 5×5. The UNet++ network used a ResNet-34 backbone network. The input was a 224×224 pixel edge image. Through four downsampling and upsampling operations, the final output was a refined edge segmentation map of the same size.

[0130] The UNet++ network structure specifically includes:

[0131] (1) Encoder part: 4 encoding blocks, each block contains 3×3 convolution, batch normalization and ReLU activation, and the number of channels is 64, 128, 256, and 512 respectively;

[0132] (2) Decoder part: 4 decoding blocks, each block contains 3×3 convolution, batch normalization, ReLU activation and 2×2 upsampling, and the number of channels is 512, 256, 128, and 64 respectively;

[0133] (3) Skip connection: adding dense connections between corresponding layers to enhance feature fusion;

[0134] (4) Output layer: 1×1 convolution maps the feature map into a binary edge map.

[0135] The network was pre-trained on a medical image segmentation dataset and then fine-tuned on 2,000 labeled cut edge images using transfer learning. The training used a combined loss function:

[0136] Loss=0.7·Dice_loss+0.3·Binary_cross_entropy

[0137] The Adam optimizer was used with an initial learning rate of 0.0001, a batch size of 8, and 100 epochs of training.

[0138] The case retrieval algorithm uses a hybrid retrieval method based on K-nearest neighbor (KNN) and cosine similarity. First, a 10-dimensional feature vector of the current edge quality data is extracted, including parameters such as average roughness, maximum roughness, edge verticality, and microcrack density. Then, the cosine similarity between this vector and the feature vectors of each case in the strategy library is calculated:

[0139] similarity(A,B)=(A·B) / (||A||·||B||)

[0140] Where A is the current feature vector and B is the feature vector of the case in the library.

[0141] The K cases with the highest similarity are selected (K is initially set to 5) and the final control strategy is determined through weighted voting, with the weights proportional to the similarity values. When the highest similarity of all candidate cases is less than 0.75, the control strategy reconstruction mechanism is triggered.

[0142] The control strategy reconstruction process uses a method that combines deep reinforcement learning with genetic algorithms. First, the high-performance strategies in the existing strategy library are regarded as the initial population. Each strategy consists of a set of control rules and parameters. Then, new candidate strategies are generated through crossover and mutation operations. The evaluation function is designed as:

[0143] F=w1·predicted_quality+w2·robustness-w3·complexity

[0144] Among them, predicted_quality is the prediction quality score, robustness is the strategy robustness score, and complexity is the strategy complexity.

[0145] The knowledge base is organized using a graph database structure, with each node representing a control strategy, and edges between nodes representing the evolution or combination of strategies. The indexing mechanism employs a multi-tiered structure: the first tier indexes by material type, the second tier indexes by cutting shape characteristics, and the third tier indexes by quality requirement level. Retrieval utilizes multi-path parallel search to improve matching efficiency. The knowledge base achieves high availability through distributed storage, and a regular compression and optimization mechanism ensures stable performance over long-term use.

[0146] like Figure 4As shown, the cutting quality assessment unit 50 includes: an image acquisition module 51 for acquiring cutting edge images; a feature extraction module 52 for applying an edge detection algorithm to extract features; a quality scoring module 53 for calculating a cutting quality score; a decision request module 54 for forming a control decision request; a strategy selection module 55 for selecting or reconstructing a strategy from a control strategy library; a scheme evaluation module 56 for evaluating a cutting control scheme; and a knowledge base management module 57 for storing and optimizing control strategies.

[0147] Step S6: cutting multiple sensor units simultaneously according to the control strategy, and performing post-cutting processing to complete the mass production of flexible tactile sensors.

[0148] In this step, a multi-station parallel cutting system is configured based on the evaluated control strategy. This modular system comprises multiple independent laser cutting workstations, each equipped with a complete multi-beam laser collaborative cutting unit and plasma assist unit. Specific cutting tasks and parameter settings are assigned to each workstation based on the control strategy. Task allocation between workstations is optimized, taking into account differences in material properties, cut shape complexity, and quality requirements. The work pace of each workstation is also coordinated to achieve continuous, streamlined production and maximize overall production capacity.

[0149] During the parallel cutting process, quality monitoring subsystems distributed across each workstation monitor the status and quality of each cutting process in real time. These subsystems collect data such as the temperature field, plasma state, and cutting front shape during the cutting process, and directly feed it back to the central control system. Based on this real-time feedback, the central control system dynamically adjusts the cutting parameters of each workstation to maintain stability and consistency during the cutting process. If anomalies in the cutting quality of a workstation are detected, the relevant parameters are immediately adjusted. In severe cases, the cutting task at that workstation is suspended until manual intervention is required to prevent the production of a large number of defective products.

[0150] This real-time monitoring and adjustment mechanism establishes a complete closed-loop process of cutting, testing, and adjusting. Within this process, the output quality of each cutting unit is evaluated in real time, and the results directly influence subsequent parameter adjustments, forming a closed-loop control system. This closed-loop process ensures that the system maintains stable cutting quality even when material properties fluctuate or environmental conditions change, achieving consistent quality throughout mass production.

[0151] After the sensor units are cut, they undergo post-cutting processing. Depending on the specific material properties and product requirements, the system may select plasma cleaning or micro-thermal treatment technologies to optimize the physical and chemical properties of the cut edges. Plasma cleaning technology uses low-temperature plasma to remove particulate residue and organic contaminants from the cut edges, improving edge cleanliness. Micro-thermal treatment technology uses precisely controlled localized heating to release residual stress introduced during the cutting process, preventing edge cracking during subsequent use. These post-processing processes are completed in specially designed workstations, forming a continuous production line with the cutting workstation.

[0152] The entire mass production process is coordinated and supervised by a full-process digital management system. This system enables data collection, analysis, and traceability from the entry of raw materials to the shipment of finished products. Each sensor unit is assigned a unique digital identifier, and all parameter settings, quality monitoring data, and post-processing records during the production process are linked to this identifier, forming a complete digital twin record. This full-process data management not only supports quality control during the production process but also provides a powerful tool for product quality traceability and problem analysis, meeting the needs of application fields with strict product traceability requirements, such as high-end medical and aerospace.

[0153] Specifically, the control strategy is followed to achieve simultaneous cutting of multiple sensor units, and post-cutting processing is performed to complete the mass production of flexible tactile sensors, including: configuring a multi-station parallel cutting system according to the control strategy to achieve simultaneous cutting of multiple sensor units; real-time monitoring of cutting quality, and establishing a cutting-detection-adjustment closed-loop process flow to ensure quality consistency in mass production.

[0154] Furthermore, the post-cutting processing process is carried out to complete the mass production of flexible tactile sensors, including: using plasma cleaning or micro-thermal treatment technology to optimize the physical and chemical properties of the cutting edges of the sensor units cut by the multi-station parallel cutting system; and realizing the full-process data collection, analysis and traceability of flexible materials through a full-process digital management system to complete the mass production of high-quality flexible tactile sensors.

[0155] like Figure 5 As shown, the batch production unit 60 includes: a multi-station parallel cutting system 61 for cutting multiple sensor units at the same time; a quality monitoring system 62 for real-time monitoring of cutting quality; a closed-loop process control system 63 for establishing a cutting-detection-adjustment closed-loop process; a post-processing process system 64 for plasma cleaning or micro-thermal treatment of the cut sensor units; and a digital management system 65 for realizing data collection and analysis throughout the entire process.

[0156] In another embodiment of the present invention, the laser cutting optimization system for mass production of the flexible tactile sensor includes: a multi-beam laser collaborative cutting unit for acquiring flexible material property data, configuring an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material property data, and generating a cutting beam and a thermal annealing beam; a plasma assist unit for generating plasma through a plasma generator and guiding the plasma to the laser cutting area to improve the laser energy absorption efficiency; a distributed state observation unit for collecting temperature field distribution, plasma density distribution, and cutting front shape data of the cutting process, and estimating cutting process state parameters based on the temperature field distribution, plasma density distribution, and cutting front shape data; a parameter collaborative optimization unit for dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma based on the cutting process state parameters; a cutting quality evaluation unit for acquiring a cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy based on the cutting quality evaluation result; and a mass production unit for realizing simultaneous cutting of multiple sensor units according to the control strategy, and performing a post-cutting processing process to complete the mass production of flexible tactile sensors.

[0157] Although specific embodiments of the present invention have been described above, these descriptions are not intended to limit the scope of protection of the present invention. Those skilled in the art may modify or replace the above embodiments without departing from the design concept of the present invention, and these modifications or equivalent replacements are all within the scope of protection of the present invention.

Claims

1. A laser cutting optimization method for mass production of flexible tactile sensors, characterized in that: include: Acquiring flexible material characteristic data, configuring an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generating a cutting beam and a thermal annealing beam; Generate plasma through a plasma generator, and guide the plasma to the laser cutting area to improve the laser energy absorption efficiency; collecting temperature field distribution, plasma density distribution and cutting front shape data during the cutting process, and estimating cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data; Dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, as well as the density and temperature of the plasma based on the cutting process state parameters; Acquire a cutting edge image, evaluate cutting quality, obtain a cutting quality evaluation result, and select or reconstruct a control strategy based on the cutting quality evaluation result; According to the control strategy, multiple sensor units are cut simultaneously, and a post-cutting processing process is performed to complete the batch production of flexible tactile sensors.

2. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 1 is characterized in that: The method of obtaining the flexible material characteristic data, configuring the ultrafast laser cutting unit and the thermal annealing laser unit according to the flexible material characteristic data, and generating the cutting beam and the thermal annealing beam includes: Identify the type of flexible materials through spectral analysis and image recognition technology, obtain the absorption, reflection and thermal conductivity characteristics of flexible materials to lasers of different wavelengths, and generate a flexible material characteristics database; Matching the optimal laser parameter combination according to the flexible material property database, configuring a femtosecond laser and a tunable wavelength continuous laser to generate the cutting beam and the thermal annealing beam; The timing, energy and spatial position of each light beam are set by a photoelectric control unit, and the optical reflector and the electrically controlled displacement platform are adjusted to achieve the coordination of the cutting light beam and the thermal annealing light beam.

3. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 2 is characterized in that: The plasma is generated by a plasma generator and directed to a laser cutting area to improve the laser energy absorption efficiency, comprising: Selecting a working gas according to the type of the flexible material, and generating plasma in the cutting area by radio frequency discharge or microwave discharge technology; The plasma is guided through a plasma transmission channel, the discharge power and the gas flow are adjusted, and the density parameters and the temperature parameters of the plasma are controlled.

4. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 3 is characterized in that: The method of guiding the plasma to the laser cutting area to improve the laser energy absorption efficiency includes: receiving plasma generated by the plasma generator, and guiding the plasma through an electromagnetic field to ensure that the plasma coincides with the laser cutting area; The interaction between the plasma and the laser in the laser cutting area is monitored to obtain energy transfer efficiency data; and based on the energy transfer efficiency data, the synergistic effect of the plasma and laser cutting is optimized.

5. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 1, characterized in that: The collecting of temperature field distribution, plasma density distribution and cutting front shape data during the cutting process, and estimating cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data, includes: Collect temperature field, plasma density and cutting front shape data during the cutting process to form a multi-dimensional sensing data set; receiving the multidimensional sensor data set, performing data preprocessing and consistency checking, and generating a valid monitoring data stream; Based on the effective monitoring data stream, a state observation equation is established using discrete-time linear time-invariant system theory, a distributed unknown input observer model is constructed, and based on the distributed unknown input observer model, the initial cutting process state parameters are estimated; Decomposing the initial cutting process state parameters into multiple subsystems, and performing parallel processing through a distributed computing architecture to estimate unknown input disturbances and obtain multi-source state estimation results; The multi-source state estimation results are weightedly fused, the reliability of the estimation results is evaluated, and the cutting process state parameters are output.

6. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 1, characterized in that: The dynamically adjusting the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, and the density and temperature of the plasma based on the cutting process state parameters includes: receiving the cutting process state parameters, comparing them with the preset cutting quality standards, and identifying adjustment parameter items; Based on the adjustment parameter items, a plasma-laser-material interaction model is established, and an optimal parameter combination is predicted through a machine learning algorithm to form a parameter adjustment instruction; According to the parameter adjustment instruction, the energy, frequency and relative position of the cutting beam and the thermal annealing beam as well as the density and temperature of the plasma are controlled to achieve real-time optimization of the cutting process.

7. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 1, characterized in that: The acquiring of the cutting edge image, evaluating the cutting quality, obtaining a cutting quality evaluation result, and selecting or reconstructing a control strategy according to the cutting quality evaluation result include: Collecting the cutting edge image, applying an edge detection algorithm to extract edge features, and obtaining edge quality data; receiving the edge quality data, comparing it with a preset standard quality template, and calculating a cutting quality score; receiving the cutting quality score, determining whether it reaches a preset quality score threshold, and forming a control decision request; Querying a preset control strategy library according to the control decision request, selecting the best matching control strategy or reconstructing a new control strategy, and generating a cutting control plan; The cutting control scheme is evaluated to obtain a cutting control scheme that passes the evaluation, and the cutting control scheme that passes the evaluation is stored in a knowledge base for continuous optimization of the control strategy library.

8. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 1, characterized in that: The method of simultaneously cutting a plurality of sensor units according to the control strategy and performing a post-cutting processing process to complete the batch production of flexible tactile sensors includes: According to the control strategy, a multi-station parallel cutting system is configured to achieve simultaneous cutting of multiple sensor units; Monitor cutting quality in real time and establish a closed-loop process of cutting-inspection-adjustment to ensure quality consistency in batch production.

9. The laser cutting optimization method for mass production of flexible tactile sensors according to claim 8, characterized in that: The post-cutting processing process is performed to complete the mass production of flexible tactile sensors, including: For the sensor unit cut by the multi-station parallel cutting system, plasma cleaning or micro-thermal treatment technology is used to optimize the physical and chemical properties of the cut edge; Through the full-process digital management system, the full-process data collection, analysis and traceability of flexible materials are realized, and the mass production of high-quality flexible tactile sensors is completed.

10. A laser cutting optimization system for mass production of flexible tactile sensors, characterized in that: include: A multi-beam laser collaborative cutting unit is used to obtain flexible material characteristic data, configure an ultrafast laser cutting unit and a thermal annealing laser unit according to the flexible material characteristic data, and generate a cutting beam and a thermal annealing beam; A plasma assist unit, configured to generate plasma through a plasma generator and guide the plasma to a laser cutting area to improve laser energy absorption efficiency; A distributed state observation unit is used to collect temperature field distribution, plasma density distribution and cutting front shape data during the cutting process, and estimate cutting process state parameters based on the temperature field distribution, plasma density distribution and cutting front shape data; a parameter collaborative optimization unit, configured to dynamically adjust the energy, frequency, and relative position of the cutting beam and the thermal annealing beam, and the density and temperature of the plasma based on the cutting process state parameters; a cutting quality assessment unit, configured to acquire a cutting edge image, assess the cutting quality, obtain a cutting quality assessment result, and select or reconstruct a control strategy based on the cutting quality assessment result; The batch production unit is used to realize the simultaneous cutting of multiple sensor units according to the control strategy, and perform post-cutting processing to complete the batch production of flexible tactile sensors.

Images