A method and device for evaluating nozzle status based on nozzle sampling detection

Abstract

The present invention belongs to the field of inkjet printing technology, and specifically relates to a nozzle state assessment method and device based on nozzle sampling detection, comprising: extracting some characteristic nozzles from the characteristic nozzles of each type of nozzle to measure the ink drop volume in the gap between substrate loading and unloading; using the sampled ink drop volume to update the parameters of the ink drop volume mixed Gaussian distribution model to obtain the ink drop volume distribution of the current characteristic nozzles of the nozzle; judging whether the current state of the nozzle is abnormal based on the current ink drop volume distribution; wherein the ink drop volume mixed Gaussian distribution model is constructed by fixing the working waveform parameters and controlling the nozzle to continuously spray ink droplets, obtaining the ink drop volume mean and standard deviation of each nozzle, and clustering all nozzles; and establishing an ink drop volume mixed Gaussian distribution model based on the volume mean corresponding to multiple characteristic nozzles selected from each type of nozzle. The present invention can improve the efficiency of nozzle state assessment during the inkjet printing production process.

Description

Technical Field

[0001] The present invention belongs to the technical field of inkjet printing, and more specifically, relates to a method and device for evaluating a printhead state based on nozzle sampling detection. Background Art

[0002] Inkjet printing is a non-contact printing technology that produces small droplets and precisely positions them, spraying a specific volume of solution onto a flexible or rigid substrate to print and prepare display screens. It is an emerging display device manufacturing technology. Compared to the traditional evaporation process used to manufacture display devices, inkjet printing offers advantages such as high material utilization, simple process, easy patterning, no need for masks, and easier manufacture of large-size panels. Currently, inkjet printing technology has become the mainstream development direction of display manufacturing, and the development of high-precision, mass-production-grade inkjet printing equipment has become a top priority for major display manufacturers and research institutions at home and abroad.

[0003] In actual production, disturbances from environmental factors can cause changes in the volume of ink droplets, which in turn causes the volume of ink in the pixel pit to not meet the requirements. Therefore, the ink droplet volume needs to be detected online during the production process to evaluate the current state of the printhead. During the production process, the beat is very tight, and the existing ink droplet volume detection technology is very time-consuming. It is impossible to detect all nozzles every time within a limited time interval. At the same time, during the printing process, since the distance between the printhead and the substrate is very close, often only 0.5mm, the ink droplet volume cannot be detected in real time. Therefore, within a limited time, quickly evaluating the current state of the printhead (whether it meets the printing requirements) based on the current ink droplet volume obtained by detection can avoid Mura defects caused by changes in ink droplet volume and improve the yield rate of the product.

[0004] Some existing methods for evaluating printhead status based on droplet volume, such as CN112757796A and CN115965911A, determine the current status by measuring droplet volume from all nozzles in flight or deposited. This is time-consuming, especially when printing on large substrates, where the number of nozzles is enormous and full inspection is impossible within the production cycle. CN115570899A provides a method for monitoring droplet volume distribution, which can extract the droplet volume of a portion of the nozzles to determine the current status of the printhead. However, for even larger numbers of nozzles, a large number of nozzles must be sampled to accurately estimate the current droplet volume distribution of the printhead. Furthermore, this method loses its effectiveness when the droplet volume distribution does not form a Gaussian distribution. In industry, currently used methods primarily determine whether a nozzle status anomaly is present by measuring film thickness during the back-end process. By the time an anomaly is detected at the back-end, production has often already continued for a considerable period, resulting in material waste. Therefore, an efficient method for evaluating printhead status in online inkjet printing manufacturing is urgently needed. Summary of the Invention

[0005] In response to the defects of the existing technology and the need for improvement, the present invention provides a nozzle status evaluation method and device based on nozzle sampling detection, which aims to improve the evaluation efficiency while ensuring high accuracy for the nozzle status evaluation task in the printing process.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for evaluating a nozzle state based on nozzle sampling detection is provided, comprising:

[0007] During the time intervals between substrate loading and unloading, some characteristic nozzles are extracted from the characteristic nozzles of each type to measure the ink droplet volume. The ink droplet volume ejected by the extracted characteristic nozzles is used to update the parameters of the pre-constructed ink droplet volume mixed Gaussian distribution model to obtain the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead. Based on the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead, whether the current state of the printhead is abnormal or not is determined, completing the printhead status assessment.

[0008] Among them, each type of nozzle and its characteristic nozzle are determined in the process of constructing the mixed Gaussian distribution model, and the construction process includes: fixing the working waveform parameters and controlling the nozzle to continuously spray ink droplets; measuring the volume of flying ink droplets sprayed by each nozzle multiple times to calculate the volume mean and volume standard deviation of the ink droplets sprayed by each nozzle to form a volume characteristic vector of the nozzle; clustering all the nozzles of the nozzle according to the volume characteristic vector of each nozzle, and selecting multiple characteristic nozzles representing the class from each type of nozzle; establishing a mixed Gaussian distribution model of ink droplet volume according to the volume mean corresponding to all the characteristic nozzles of the nozzle, as the ink droplet volume distribution of the ink droplets sprayed by all the characteristic nozzles of the nozzle under fixed working waveform parameters.

[0009] Furthermore, the clustering is implemented by:

[0010] S1, preset number of clusters K;

[0011] S2. Randomly select K nozzles from all nozzles in the nozzle, and use the volume feature vectors of these K nozzles as the initial mean volume feature vectors of the K clusters.

[0012] S3. Calculate the distance between the volume feature vector of each remaining nozzle and the mean volume feature vector of the K clusters, and divide each remaining nozzle into the cluster closest to it; calculate the mean of the volume feature vectors of each nozzle in each cluster after division, and use it as the new mean volume feature vector of the cluster;

[0013] S4. Calculate the distance between the volume feature vector of each nozzle in all nozzles of the nozzle and the new mean volume feature vector of the K clusters, re-execute the cluster division operation for each nozzle in S3 and the update operation of the mean volume feature vector of each cluster, and repeat this step until the mean volume feature vector of each cluster does not change.

[0014] Furthermore, the clustering implementation method further includes:

[0015] S5. Calculate the sum of the squares of the distances between the volume feature vector of each nozzle and the mean volume feature vector of the cluster to which it belongs;

[0016] S6. Increase the K value and repeat S1 until the sum converges. Use the current K value as the number of clusters for nozzle clustering to complete clustering of all nozzles of the nozzle.

[0017] Furthermore, the characteristic nozzles selected from each type of nozzles include: the nozzles with the largest volume mean, the nozzles with the smallest volume mean, the nozzles with the largest volume standard deviation, and the nozzles with the smallest volume standard deviation in the nozzles of the type.

[0018] Furthermore, the characteristic nozzles also include nozzles selected by Latin cube sampling, which is specifically implemented as follows:

[0019] The volume mean set and volume standard deviation set of all nozzles in each type of nozzle are divided into a intervals, and the sample extraction probability of each interval is the same;

[0020] A sample is randomly selected from each interval, and samples extracted from the volume mean and volume standard deviation are randomly combined in pairs to construct a volume feature vector as a set of candidate volume feature vectors, where the value of a is not less than 20% of the total number of nozzles of this type;

[0021] The constructed volume feature vectors that exist in the volume feature vectors corresponding to the nozzle of this type are retained from the candidate volume feature vector set; the constructed volume feature vectors that do not exist in the volume feature vectors corresponding to the nozzle of this type are deleted from the candidate volume feature vector set, and the volume feature vector closest to the volume feature vector is selected from the volume feature vectors corresponding to the nozzle of this type and added to the candidate volume feature vector set to obtain an updated candidate volume feature vector set;

[0022] The nozzles corresponding to a volume feature vectors in the updated candidate volume feature vector set are respectively used as characteristic nozzles of this type of nozzles.

[0023] Furthermore, the mixed Gaussian distribution model is expressed as:

[0024]

[0025] Where, β m is the coefficient of the mth Gaussian distribution model, M represents the total number of Gaussian distribution sub-models in the mixed Gaussian distribution; φ(μ v ∣θ m ) is the probability density of the mth Gaussian distribution model, μ v represents the mean volume of ink droplets of characteristic nozzles; θ m =(μ m ,σ m 2 ), μ m , σ m are the mean and standard deviation of the mth Gaussian distribution model respectively;

[0026] The parameters of the pre-built ink drop volume mixed Gaussian distribution model are updated in the following way:

[0027] The parameter β of the ink drop volume mixed Gaussian distribution model is used m and φ(μ v ∣θ m ), calculate the volume mean V of the jth characteristic nozzle inspected by the mth Gaussian distribution model j Responsiveness

[0028]

[0029] According to the responsiveness, the parameters of the new ink drop volume mixed Gaussian distribution model are calculated:

[0030]

[0031]

[0032]

[0033] use and Update the responsiveness The corresponding μ in the function m , σ m 2 and β m Take the value and repeatedly calculate the volume mean V of the characteristic nozzle holes sampled j Responsiveness Until Convergence, complete parameter update, after convergence The corresponding ink drop volume mixed Gaussian distribution model is used as the ink drop volume distribution of ink drops ejected by all characteristic nozzles in the current state of the printhead.

[0034] Furthermore, if the ink drop volume distribution satisfies:

[0035] μ mmax +3σ mmax ≤V s +b%;

[0036] μ mmin +3σ mmin ≥V s -b%;

[0037] The current printhead is considered to be working normally; otherwise, the printhead is considered to be abnormal;

[0038] Where μ mmax 、μ mmin They represent the maximum volume mean and the minimum volume mean in the droplet volume distribution, σ mmax , σ mmin Represent the maximum volume standard deviation and the minimum volume standard deviation in the droplet volume distribution, V s It represents the set value of the volume of ink droplets ejected by the nozzle, and b% represents the amplitude of the set volume of ink droplets ejected by the nozzle.

[0039] The present invention also provides a nozzle state evaluation device based on nozzle sampling detection, which is used to execute the nozzle state evaluation method described above. The device includes: a motion module, a vision module, and a sampling detection and evaluation module;

[0040] Among them, the motion module is used to control the movement of the nozzle and the vision module; the vision module is used to collect the volume of ink droplets; the sampling detection and evaluation module is used to construct the ink droplet volume distribution of ink droplets sprayed by all characteristic nozzles of the nozzle, and evaluate the current state of the nozzle based on the ink droplet volume distribution of ink droplets sprayed by all characteristic nozzles in the current state of the nozzle.

[0041] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0042] (1) In view of the problem of low efficiency in ink droplet volume detection in the task of nozzle status assessment in the production process, the present invention provides a sampling detection method. First, based on the volume mean, volume standard deviation and other data of the ink droplets ejected by the nozzles under fixed working parameters, the nozzles are clustered, which can effectively classify nozzles with similar characteristics into one category. For nozzles of the same type, it is only necessary to detect the volume of ink droplets ejected by characteristic nozzles that can represent the current state of this type of nozzles, and construct the ink droplet volume distribution of the ink droplets ejected by the characteristic nozzles under fixed working parameters. It is not necessary to detect the ink droplets of all nozzles of the nozzle, which effectively reduces the number of nozzles to be detected. Furthermore, when performing actual printhead status assessment, only the droplet volumes of a subset of characteristic nozzles need to be sampled, rather than all of them. By updating the parameters of the droplet volume distribution of the droplets ejected by these characteristic nozzles, the droplet volume distribution of all characteristic nozzles in the current state of the printhead can be obtained. This method, which further reduces the number of nozzles sampled based on the characteristic nozzles, can further reduce the number of nozzles tested within the loading and unloading time interval. This method can still meet efficiency requirements when printing large-scale substrates requiring tens of thousands of nozzles. Furthermore, the droplet volume distribution constructed by this method is a mixed Gaussian distribution model, which theoretically can effectively fit various forms of nozzle volume distributions, even if the droplet volume distribution is not Gaussian. This model is highly universal and can guarantee assessment accuracy. Therefore, compared to existing methods, the present invention can sample and test a small number of nozzles within a limited time interval to complete the entire printhead status assessment, while maintaining detection accuracy and meeting production cycle requirements.

[0043] (2) When clustering all nozzle holes, the present invention proposes a clustering method based on K value selection, which classifies similar nozzle holes into one category as much as possible, reduces the error in nozzle clustering, and ensures the accuracy of detection.

[0044] (3) The present invention proposes a method for selecting characteristic nozzles, which symbolically selects the nozzles corresponding to the maximum and minimum volume mean values ​​and the nozzles corresponding to the maximum and minimum volume standard deviations in each type of nozzles. In addition, a cubic sampling method is used to randomly select characteristic nozzles, so that the selected nozzles represent the volume characteristics of this type of nozzles to the greatest extent, thereby improving the accuracy of the method of the present invention in constructing the ink droplet volume distribution and further ensuring the accuracy of the nozzle status assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of a method for evaluating a nozzle state based on nozzle sampling detection provided by an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of nozzle clustering provided in an embodiment of the present invention;

[0047] Figure 3A schematic diagram of ink droplet volume distribution construction provided by an embodiment of the present invention;

[0048] Figure 4 A schematic diagram of nozzle status determination provided by an embodiment of the present invention;

[0049] Figure 5 A schematic diagram of a nozzle status assessment device based on nozzle sampling detection provided by an embodiment of the present invention.

[0050] Throughout the drawings, the same reference numerals are used to denote the same elements or structures, wherein:

[0051] 1 is the nozzle module, 2 is the motion module, 31 is the light source, 32 is the ink drop observation camera, and 4 is the sampling detection and evaluation module. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0053] Example 1

[0054] A method for evaluating a nozzle state based on nozzle sampling detection, comprising:

[0055] During the time intervals between substrate loading and unloading, some characteristic nozzles are extracted from the characteristic nozzles of each type to measure the ink droplet volume. The ink droplet volume ejected by the extracted characteristic nozzles is used to update the parameters of the pre-constructed ink droplet volume mixed Gaussian distribution model to obtain the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead. Based on the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead, whether the current state of the printhead is abnormal or not is determined, completing the printhead status assessment.

[0056] Among them, each type of nozzle and its characteristic nozzle are determined in the process of constructing the mixed Gaussian distribution model, and the above-mentioned construction process includes: fixing the working waveform parameters and controlling the nozzle to continuously spray ink droplets; measuring the volume of flying ink droplets sprayed by each nozzle multiple times to calculate the volume mean and volume standard deviation of the ink droplets sprayed by each nozzle to form the volume characteristic vector of the nozzle; clustering all the nozzles of the nozzle according to the volume characteristic vector of each nozzle, and selecting multiple characteristic nozzles representing the class from each type of nozzle; establishing a mixed Gaussian distribution model of ink droplet volume according to the volume mean corresponding to all the characteristic nozzles of the nozzle, as the ink droplet volume distribution of the ink droplets sprayed by all the characteristic nozzles of the nozzle under fixed working waveform parameters.

[0057] The nozzle status evaluation method provided in this embodiment is generally divided into a model building stage, a sampling detection and evaluation stage, such as Figure 1 shown.

[0058] Model building phase:

[0059] For a specific period of time, with fixed operating waveform parameters, the printhead continuously ejects ink droplets. The droplet volume ejected from each nozzle is recorded, and the mean and standard deviation of the droplet volume ejected from each nozzle are calculated to form the volume feature vector for that nozzle. Based on the volume feature vectors of each nozzle, all nozzles in the printhead are clustered, and a subset of nozzles in each cluster are extracted as characteristic nozzles. Based on the volume mean data of the characteristic nozzles, a characteristic nozzle droplet volume distribution model is established (i.e., the droplet volume distribution of the ink droplets ejected from all characteristic nozzles of the printhead under fixed operating waveform parameters).

[0060] Sampling, testing and evaluation stage:

[0061] During loading and unloading, a certain number of characteristic nozzles are sampled for droplet volume measurement to update the characteristic nozzle droplet volume distribution model. The characteristic parameters such as the droplet volume mean and standard deviation in the updated droplet volume distribution model can be compared with the droplet volume standards required during the actual printing process. If the standards are met, the current printhead status is considered normal and the printing task can continue. Otherwise, the machine needs to be stopped for inspection.

[0062] The inkjet printing sampling detection method provided by this method can obtain the current status of the printhead by sampling the volume of ink droplets ejected from a small number of nozzles during the printing process. It is suitable for applications such as manufacturing high-resolution displays and electronic components using inkjet printing.

[0063] It should be noted that in the actual printing process, the volume of the ink droplets ejected will change due to external factors such as device aging caused by continuous operation of the nozzle, changes in ink viscosity and environmental changes. Due to factors such as processing errors and nozzle structure layout, the mean and standard deviation of the volume distribution of the ink droplets ejected by each nozzle are not exactly the same, and they all obey the normal distribution X~N(μ,σ 2 ). Therefore, it is preferable to construct a data set in the following way: fix the working waveform parameters, use the nozzle to continuously eject ink droplets, measure the volume of ink droplets in all nozzle holes multiple times, and obtain multiple sets of ink droplet volume data for each nozzle hole. Based on the obtained ink droplet volume data, calculate the mean ink droplet volume μ for each nozzle hole v , standard deviation σ v The final dataset D can be expressed as {D=(i,μ vi ,σ vi )} i＝1,2...n , where i represents the nozzle number, μ virepresents the mean value of the ink drop volume of the i-th nozzle, σ vi represents the standard deviation of the droplet volume of the i-th nozzle. The longer the volume data is collected, the better. During the data collection period, the nozzle is kept in continuous operation to simulate the continuous operation of the nozzle during the actual printing process.

[0064] Some nozzles have similar characteristics, so they can be considered as a type of nozzle during inspection. That is, by inspecting some of the nozzles in this type, the status of this type of nozzle can be judged, which can effectively reduce the number of nozzle inspections.

[0065] As a preferred embodiment, Figure 2 As shown, the implementation of the above clustering includes:

[0066] S1, preset number of clusters K;

[0067] S2. Randomly select K nozzles from all nozzles in the nozzle, and use the volume feature vectors of these K nozzles as the initial mean volume feature vectors of the K clusters.

[0068] S3. Calculate the distance between the volume feature vector of each remaining nozzle and the mean volume feature vector of the K clusters, and divide each remaining nozzle into the cluster closest to it; calculate the mean of the volume feature vectors of each nozzle in each cluster after division, and use it as the new mean volume feature vector of the cluster;

[0069] S4. Calculate the distance between the volume feature vector of each nozzle in the nozzle and the new mean volume feature vector of the K clusters, re-execute the cluster division operation for each nozzle in S3 and the update operation of the mean volume feature vector of each cluster, and repeat this step until the mean volume feature vector of each cluster does not change.

[0070] Select an appropriate value for the number of nozzle clustering clusters K. If the K value is too large, there will be too many nozzle classification clusters, the number of characteristic nozzles will increase, and the detection time will be greatly increased. If the K value is too small, the nozzle classification will not be thorough, and the extracted nozzles will not represent the characteristics of this type of nozzle. As a preferred implementation method, the clustering implementation method also includes:

[0071] S5. Calculate the sum of the squares of the distances between the volume feature vector of each nozzle and the mean volume feature vector of the cluster to which it belongs;

[0072] S6. Increase the K value and repeat S1 until the above sum converges. Use the current K value as the number of clusters for nozzle clustering to complete the clustering of all nozzles of the nozzle.

[0073] After clustering the nozzles, a subset of nozzles are extracted from each cluster as characteristic nozzles. The ink droplet volume status is then measured to determine the current printhead status. In a preferred embodiment, the characteristic nozzles selected from each cluster include: the nozzles with the largest mean volume, the nozzles with the smallest mean volume, the nozzles with the largest volume standard deviation, and the nozzles with the smallest volume standard deviation within that cluster.

[0074] As a preferred embodiment, the characteristic nozzles further include nozzles selected by Latin cube sampling, and the specific implementation is as follows:

[0075] The volume mean set and volume standard deviation set of all nozzles in each type of nozzle are divided into a intervals, and the sample extraction probability of each interval is the same. The more concentrated the sample area, the smaller the interval;

[0076] A sample is randomly selected from each interval, and each interval is selected only once. The samples extracted from the volume mean and volume standard deviation are randomly combined in pairs to construct a volume feature vector as the candidate volume feature vector set, where the value of a is not less than 20% of the total number of nozzles of this type;

[0077] The constructed volume feature vectors that exist in the volume feature vectors corresponding to this type of nozzle are retained from the candidate volume feature vector set; the constructed volume feature vectors that do not exist in the volume feature vectors corresponding to this type of nozzle are deleted from the candidate volume feature vector set, and the volume feature vector closest to the volume feature vector is selected from the volume feature vectors corresponding to this type of nozzle and added to the candidate volume feature vector set to obtain an updated candidate volume feature vector set; the nozzles corresponding to a volume feature vectors in the updated candidate volume feature vector set are respectively used as characteristic nozzles of this type of nozzle, so that the extracted characteristic nozzles can be more evenly distributed in each type, and the conditions of each type of nozzle can be better reflected.

[0078] When printing on large-sized substrates, multiple print heads and tens of thousands of nozzles are used. It is still unrealistic to detect the ink drop volume of all characteristic nozzles within the loading and unloading time. Figure 3 As shown in the figure, a distribution model can be established for the characteristic nozzles. By extracting a certain number of characteristic nozzles to measure the ink drop volume, the characteristic nozzle drop volume distribution model is updated to determine the current state of the printhead. Theoretically, the characteristic nozzle drop volume distribution can be any distribution, so the traditional single Gaussian distribution X~N(μ,σ 2 ), does not necessarily meet the fitting requirements. The mixed Gaussian distribution is composed of multiple Gaussian distributions and can theoretically fit any distribution. Therefore, it is preferable to use the mean volume μ of the characteristic nozzle in the data set. v, a mixed Gaussian model describing the volume distribution of ink droplets from characteristic nozzles is constructed, which is expressed as:

[0079]

[0080] Where, β m is the coefficient of the mth Gaussian distribution sub-model, representing the proportion of the distribution occupied by the mth Gaussian distribution sub-model, M represents the total number of Gaussian distribution sub-models in the mixed Gaussian distribution; φ(μ v ∣θ m ) is the probability density of the mth Gaussian distribution model, μ v represents the mean volume of ink droplets of characteristic nozzles; θ m =(μ m ,σ m 2 ), μ m , σ m are the mean and standard deviation of the mth Gaussian distribution model respectively;

[0081] The parameters of the pre-built mixed Gaussian distribution model of the ink droplet volume ejected by all characteristic nozzles of the printhead are updated in the following manner:

[0082] Introducing hidden variables γ jm , which is the response of the observed sample nozzle volume V to the mth sub-model, and the logarithmic likelihood function of the current data feature nozzle droplet volume is obtained

[0083]

[0084] Determine the Q function

[0085] Q(θ,θ (i) )=E[logP(V,γ|θ)|V,θ (i) ]

[0086]

[0087] Based on the established characteristic nozzle droplet volume distribution model parameter β m ,φ(μ v ∣θ m ), calculate the mth Gaussian distribution in the ink drop volume distribution model for the current observed sample nozzle volume V j Responsiveness

[0088]

[0089] Will Substitute Q(θ,θ (i) ),available

[0090]

[0091] Calculate the parameters of the new iterative ink drop volume distribution model, that is, find the maximum value of the Q function with respect to θ. Just need to Q(θ,θ (i) )Find the partial derivative and set it to 0. exist Under these conditions, find the partial derivative and set it to 0:

[0092]

[0093]

[0094]

[0095] Repeat the above calculation until Converge and obtain the characteristic nozzle droplet volume distribution model in the current state.

[0096] It should be noted that the parameter update may also adopt Bayesian estimation or Bootstrap algorithm.

[0097] As a preferred embodiment, Figure 4 As shown, the method for judging the state of the ink droplet volume of the nozzle is: during the inkjet printing process, the change in the volume of the ink droplet ejected from the nozzle is required not to exceed ±b% of the set value.

[0098] For Gaussian distribution, the samples are basically distributed within three standard deviations of the mean, that is, (μ-3σ,μ+3σ). Therefore, if the posterior distribution satisfies:

[0099] μ mmax +3σ mmax ≤V s +b%;

[0100] μ mmin +3σ mmin ≥V s -b%;

[0101] The current printhead is considered to be working normally; otherwise, the printhead is considered to be abnormal;

[0102] Where μ mmax 、μ mmin They represent the maximum volume mean and the minimum volume mean in the droplet volume distribution, σ mmax , σ mmin Represent the maximum volume standard deviation and the minimum volume standard deviation in the droplet volume distribution, V s It represents the set value of the volume of ink droplets ejected by the nozzle, and b% represents the amplitude of the set volume of ink droplets ejected by the nozzle.

[0103] Example 2

[0104] A nozzle state evaluation device based on nozzle sampling detection is used to execute the nozzle state evaluation method described above, the device includes: a motion module, a visual module, a sampling detection and evaluation module; wherein, Figure 5 As shown, the motion module 2 is used to control the movement of the nozzle module 1 and the vision module; the vision module is used to collect the volume of ink droplets; the sampling detection and evaluation module 4 is used to construct the ink droplet volume distribution of the ink droplets ejected by all the characteristic nozzles of the nozzle, and evaluate the current state of the nozzle based on the ink droplet volume distribution of the ink droplets ejected by all the current characteristic nozzles of the nozzle.

[0105] The motion module 2 moves the printhead module 1 to the ink drop volume measurement area. The printhead module 1 reads the current working waveform parameters from the sampling detection and evaluation module 4, reads the sequence number of the characteristic nozzle hole sampled from the sampling detection and evaluation module 4, and generates ink drops.

[0106] The vision module includes a light source 31 and an ink droplet observation camera 32. The lenses of the light source 31 and the ink droplet observation camera 32 are coaxially mounted on both sides. They are used to collect ink droplet images generated by the printhead module 1 and transmit them to the sampling inspection and evaluation module 4. After image processing, the current ink droplet volume is obtained.

[0107] The random inspection and evaluation module 4 reads the collected ink drop volume information and executes the nozzle status evaluation method as described in the first embodiment.

[0108] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A nozzle status assessment method based on nozzle sampling detection, characterized in that: include: During the time intervals between substrate loading and unloading, some characteristic nozzles are extracted from the characteristic nozzles of each type to measure the ink droplet volume. The ink droplet volume ejected by the extracted characteristic nozzles is used to update the parameters of the pre-constructed ink droplet volume mixed Gaussian distribution model to obtain the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead. Based on the ink droplet volume distribution of the ink droplets ejected by all characteristic nozzles in the current state of the printhead, whether the current state of the printhead is abnormal or not is determined, completing the printhead status assessment. Each type of nozzle and its characteristic nozzle are determined during the construction of the mixed Gaussian distribution model. The construction process includes: fixing the operating waveform parameters and controlling the nozzle to continuously eject ink droplets; repeatedly measuring the volume of flying ink droplets ejected by each nozzle to calculate the volume mean and volume standard deviation of the ink droplets ejected by each nozzle to form a volume feature vector of the nozzle; clustering all the nozzles of the nozzle based on the volume feature vectors of each nozzle, and selecting multiple characteristic nozzles representing each type of nozzle; and establishing a mixed Gaussian distribution model of ink droplet volume based on the volume mean corresponding to all characteristic nozzles of the nozzle, as the ink droplet volume distribution of ink droplets ejected by all characteristic nozzles of the nozzle under fixed operating waveform parameters. The clustering is implemented by: S1, preset number of clusters K; S2. Randomly select K nozzles from all nozzles in the nozzle, and use the volume feature vectors of these K nozzles as the initial mean volume feature vectors of the K clusters. S3. Calculate the distance between the volume feature vector of each remaining nozzle and the mean volume feature vector of the K clusters, and divide each remaining nozzle into the cluster closest to it; calculate the mean of the volume feature vectors of each nozzle in each cluster after division, and use it as the new mean volume feature vector of the cluster; S4. Calculate the distance between the volume feature vector of each nozzle in all nozzles of the nozzle and the new mean volume feature vector of the K clusters, re-execute the cluster division operation for each nozzle in S3 and the update operation of the mean volume feature vector of each cluster, and repeat this step until the mean volume feature vector of each cluster does not change.

2. The nozzle status evaluation method according to claim 1, characterized in that: The clustering implementation also includes: S5. Calculate the sum of the squares of the distances between the volume feature vector of each nozzle and the mean volume feature vector of the cluster to which it belongs; S6. Increase the K value and repeat S1 until the sum converges. Use the current K value as the number of clusters for nozzle clustering to complete clustering of all nozzles of the nozzle.

3. The nozzle status evaluation method according to claim 1, characterized in that: The characteristic nozzles selected from each type of nozzles include: the nozzle with the largest volume mean, the nozzle with the smallest volume mean, the nozzle with the largest volume standard deviation, and the nozzle with the smallest volume standard deviation in the nozzle type.

4. The nozzle status evaluation method according to claim 3, characterized in that: The characteristic nozzles also include nozzles selected by Latin cube sampling, which is specifically implemented as follows: The volume mean set and volume standard deviation set of all nozzles in each type of nozzle are divided into a intervals, and the sample extraction probability of each interval is the same; A sample is randomly selected from each interval, and samples extracted from the volume mean and volume standard deviation are randomly combined in pairs to construct a volume feature vector as a set of candidate volume feature vectors, where the value of a is not less than 20% of the total number of nozzles of this type; The constructed volume feature vectors that exist in the volume feature vectors corresponding to the nozzle of this type are retained from the candidate volume feature vector set; the constructed volume feature vectors that do not exist in the volume feature vectors corresponding to the nozzle of this type are deleted from the candidate volume feature vector set, and the volume feature vector closest to the volume feature vector is selected from the volume feature vectors corresponding to the nozzle of this type and added to the candidate volume feature vector set to obtain an updated candidate volume feature vector set; The nozzles corresponding to a volume feature vectors in the updated candidate volume feature vector set are respectively used as characteristic nozzles of this type of nozzles.

5. The nozzle status evaluation method according to claim 1, characterized in that: The mixed Gaussian distribution model is expressed as: Where, β m is the coefficient of the mth Gaussian distribution model, M represents the total number of Gaussian distribution sub-models in the mixed Gaussian distribution; φ(μ v ∣θ m ) is the probability density of the mth Gaussian distribution model, μ v represents the mean volume of ink droplets ejected by the characteristic nozzle; θ m =(μ m ,σ m 2 ), μ m , σ m are the mean and standard deviation of the mth Gaussian distribution model respectively; The parameters of the pre-built ink drop volume mixed Gaussian distribution model are updated in the following way: The parameter β of the ink drop volume mixed Gaussian distribution model is used m and φ(μ v ∣θ m ), calculate the volume mean V of the jth characteristic nozzle inspected by the mth Gaussian distribution model j Responsiveness According to the responsiveness, the parameters of the new ink drop volume mixed Gaussian distribution model are calculated: use and Update the responsiveness The corresponding μ in the function m , σ m 2 and β m Take the value and repeatedly calculate the volume mean V of the characteristic nozzle holes sampled j Responsiveness Until Convergence, complete parameter update, after convergence The corresponding ink drop volume mixed Gaussian distribution model is used as the ink drop volume distribution of ink drops ejected by all characteristic nozzles in the current state of the printhead.

6. The method for evaluating the state of a printhead according to any one of claims 1 to 5, characterized in that: If the droplet volume distribution satisfies: m mmax +3s mmax ≤V s +b%; m mmin +3s mmin ≥V s -b%; The current printhead is considered to be working normally; otherwise, the printhead is considered to be abnormal; Where μ mmax 、μ mmin They represent the maximum volume mean and the minimum volume mean in the droplet volume distribution, σ mmax , σ mmin Represent the maximum volume standard deviation and the minimum volume standard deviation in the droplet volume distribution, V s It represents the set value of the volume of ink droplets ejected by the nozzle, and b% represents the amplitude of the set volume of ink droplets ejected by the nozzle.

7. A nozzle status assessment device based on nozzle sampling detection, characterized in that: For executing the nozzle status assessment method according to any one of claims 1 to 6, the device comprises: a motion module, a vision module, and a sampling detection and assessment module; Among them, the motion module is used to control the movement of the nozzle and the vision module; the vision module is used to collect the volume of ink droplets; the sampling detection and evaluation module is used to construct the ink droplet volume distribution of ink droplets sprayed by all characteristic nozzles of the nozzle, and evaluate the current state of the nozzle based on the ink droplet volume distribution of ink droplets sprayed by all characteristic nozzles in the current state of the nozzle.

Images