Multimodal printer state prediction method and system

Abstract

The application discloses a multi-modal printer state prediction method and system, and relates to the technical field of state prediction.The application can identify a state segment at risk in advance before an obvious fault occurs by introducing a state evolution trend analysis mechanism, early warning and trend guidance of key parameter changes in a printing task process are realized, and the risk of printing quality decline or equipment damage caused by state mutation is effectively reduced.Meanwhile, the application adopts a multi-modal data-driven conditional modeling and state clustering mode, adaptability and generalization ability of the prediction model to complex operating conditions are improved, and the application is especially suitable for practical industrial environments such as frequent switching of printing tasks and variable equipment operating states.

Description

Technical Field

[0001] This invention relates to the field of state prediction technology, and in particular to a method and system for predicting the state of a multimodal printer. Background Technology

[0002] With the continuous evolution of smart manufacturing and industrial IoT technologies, printing equipment is gradually transforming from traditional single-function systems to multimodal intelligent systems. Accurate prediction and anomaly warning of equipment operating status have become crucial for improving printing task stability and print quality. Especially given the increasingly complex structure of printers and the dynamically changing working environment, relying solely on a single data dimension (such as temperature, voltage, number of pages printed) is insufficient to comprehensively control the equipment's operating status. Therefore, the need for integrating multimodal data (structural sensor signals, logs, print load, etc.) and using intelligent algorithms to predict state trends is becoming increasingly prominent. However, existing technologies mostly focus on assessing the current state of the equipment or correcting anomalies based on static models. There is a lack of a systematic method for trend identification, potential anomaly segment discrimination, and state evolution prediction for different stages of dynamic evolution. Accurate modeling of equipment state evolution patterns remains a significant challenge, particularly in scenarios such as cold starts and frequent mode switching.

[0003] CN118963684B discloses a printer state correction method and system. This method establishes a theoretical aging curve model between aging parameters and theoretical data curves to classify and determine the actual printing effect level. Based on the curve change level and aging level, it determines the attenuation coefficient to correct the current printer state and trigger the state correction process. Although this method can compensate for the aging state of multiple sensors to some extent, thereby improving the accuracy of the correction coefficient, its core is still based on fitting and correcting a single indicator using a static aging model. It fails to capture the complex dynamic correlations between multimodal data, especially lacking a response mechanism for data trend changes and the ability to predict future state evolution trends. This makes it difficult to adapt to complex operating scenarios where equipment modes frequently switch or early potential anomalies gradually accumulate.

[0004] CN119766019A proposes a model predictive control method and system for industrial-grade 3D printers. By constructing a model predictive controller (MPC) and combining motor state information to output the optimal voltage vector in real time to control the motor's operating state, it improves motor response speed and printing accuracy. This method provides an optimized path control strategy for problems such as path tracking and torque control during 3D printing. However, its core focus is on motor-level motion control optimization, primarily addressing hardware response issues caused by the complexity of the printing path. It does not involve the evolutionary modeling and prediction of the printer's overall system state, nor does it address how to mine potential abnormal states or trend changes from multimodal data. Its ability to support the full-cycle state progression and prediction of equipment in complex scenarios is relatively limited.

[0005] In summary, existing printer state correction or control methods mostly focus on single-modal data modeling or local control optimization of unit hardware, and have not yet formed a complete framework for fusing multimodal data and establishing an evolution trend prediction model. In particular, they lack accurate state identification and prediction mechanisms under cold start and sudden modal fluctuation conditions. Summary of the Invention

[0006] In view of the problems with existing printer correction or control methods under cold start and sudden modal fluctuation conditions, this invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to achieve dynamic evolution prediction of printing equipment status and early identification of potential faults based on feature vector-based conditional segmentation modeling and trend prediction mechanism, combined with abnormal segment temporal clustering and state-level transition sequence generation technology.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] In a first aspect, the present invention provides a method for predicting the state of a multimodal printer, comprising: acquiring a multimodal data stream set, determining whether it is in the cold start stage, and outputting a stage label; conditionally segmenting the data stream set according to the stage label, and calculating feature vectors for each segment; constructing a local polynomial regression model for each segment, and mapping the trend vectors of key indicators within the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, marking the current segment as a potential abnormal segment; performing time-dimensional clustering and sorting on the set of potential abnormal segments, generating a state-level transition sequence through inter-segment transition associations, and outputting the state prediction result based on trend consistency; and adjusting the weights of each regression model based on the comparison between the state prediction result and the subsequent real state label.

[0010] As a preferred embodiment of the multimodal printer state prediction method of the present invention, the determination of the cold start stage includes: constructing a window for the first k time slices in the multimodal data stream set, extracting the modal feature component set, and calculating the heteroscedasticity for each modal feature to form a feature heteroscedasticity vector, where k is a constant; calculating the fluctuation density within the current window based on the feature heteroscedasticity vector, and calculating the ratio of the average fluctuation density stored in the historical stable stage as a cold start discrimination factor; if the ratio is greater than or equal to the density offset threshold, the output stage label is cold start; otherwise, the output stage label is stable stage.

[0011] As a preferred embodiment of the multimodal printer state prediction method of the present invention, the conditional segmentation includes: mapping stage labels to multimodal data stream sets; for each stage label, extracting the data stream sequence generated in the corresponding stage; dividing the data stream sequence into several continuous data segments based on the time order between data points, with each segment having a fixed period length; and in each segment, further segmenting the data points within the segment based on whether the printhead temperature exceeds the upper or lower fluctuation range of the average value of the corresponding stage, thereby generating a conditional segment set.

[0012] As a preferred embodiment of the multimodal printer state prediction method of the present invention, the construction of the local polynomial regression model includes: for any feature vector, retrieving the last T in the corresponding conditional slice. e A sequence of key indicators per unit time; where key indicators include, but are not limited to, modal values ​​of printhead temperature and feed rate; a weight sequence is introduced onto the key indicator sequence, where each weight is dynamically calculated based on the modal variance and historical mean of that data point, so that high fluctuation points account for a higher proportion in the fitting, and a weighted vectorized trend sequence s is constructed. k Based on s k The key indicator sequence is constructed using a polynomial fitting modeling method of no more than order 3, and a local prediction expression is independently constructed for each modal indicator.

[0013] In a preferred embodiment of the multimodal printer state prediction method of the present invention, the confirmation of the abnormal segment includes: calculating a continuous difference Δ on the trend vector. k And construct a trend mutation factor group, where any trend mutation factor is |Δ j+1 -Δ j |; Combine all trend-changing factor groups in the historical segments of each stage to calculate the stability reference interval. in, and θ represents the mean and standard deviation of the mutation factor, respectively, and θ is the adjustment threshold coefficient. The system determines whether any mutation factor in the trend mutation factor group exceeds the stability reference interval Λ. If it does, the current segment is identified as a deviation segment, and the trigger point and corresponding time point are recorded. If it is a deviation segment, the system extracts the multimodal data set contained in the feature vector of the segment, calculates the variation amplitude within the current window for each mode, and constructs a modal fluctuation sequence. If the variation amplitude of at least two items in the modal fluctuation sequence is higher than the upper bound of the historical modal interval, the segment is confirmed as an anomalous segment.

[0014] As a preferred embodiment of the multimodal printer state prediction method of the present invention, the generation of the state level transition sequence includes: constructing an indexed linked list of monotonically increasing segments from the set of abnormal segments according to timestamp information, and embedding the inter-segment time interval parameter γ in the indexed linked list structure. tEstablish a temporal sequence constraint structure; group adjacent potential anomalous segments in γ t Fragments within the same time frame are grouped into clusters, denoted as TQ, and each TQ is assigned a fragment sequence label, forming a fragment cluster sequence aggregated along the time dimension; based on the consistency ratio η of fragments within each TQ in the characteristic trend direction... h Construct a cross-cluster transfer criterion, the criterion satisfying η h When the state transition path from TQ to TQ+1 is greater than the preset directional consistency threshold, a state transition path is established. Based on the state transition path, a state-level transition sequence is generated, and each pair of consecutive cluster associations in the transition sequence is marked as a first-level state transition pair.

[0015] As a preferred embodiment of the multimodal printer state prediction method of the present invention, the state prediction result based on trend consistency includes: for each pair of continuous segment clusters, comparing whether the overall trend direction of the corresponding segments in the original sequence remains continuous; when trend consistency is established, it is considered that the state marked by the previous segment cluster remains unchanged in the subsequent segment cluster, and the state label of the previous segment cluster is continued to be assigned to the subsequent segment cluster, which is regarded as a continuous evolution process of the state; if it is determined that the trend is inconsistent, it is regarded as a potential transition of the state, and the state label of the previous segment cluster is not directly continued, the current segment cluster is independently labeled, and the position of the segment cluster is marked as a state change node in the state level transition sequence; after completing the trend consistency judgment and state label processing of all continuous segment clusters, the final state label of each segment cluster is rearranged according to the original time order, and a state prediction result sequence covering all segment clusters is output as the final trend-guided state prediction output.

[0016] Secondly, the present invention provides a multimodal printer status prediction system, which includes: a cold start discrimination module, used to acquire a multimodal data stream set, determine whether it is in the cold start stage, and output a stage label;

[0017] The conditional sharding module is used to conditionally shard the data stream collection based on the stage label and calculate the feature vector for each shard.

[0018] The anomaly labeling module is used to construct a local multinomial regression model for each segment, and to map the trend vector of key indicators in the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, the current segment is labeled as a potential anomaly segment.

[0019] The state transition module is used to cluster and sort the set of potential abnormal segments in the time dimension, generate a state-level transition sequence through the transition association between segments, and output the state prediction result based on trend consistency.

[0020] The model adjustment module is used to adjust the weights of each regression model based on the comparison between the state prediction results and the subsequent real state labels.

[0021] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program instructions, when executed by the processor, implement the steps of the multimodal printer state prediction method as described in the first aspect of the present invention.

[0022] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the multimodal printer state prediction method as described in the first aspect of the present invention.

[0023] The beneficial effects of this invention are as follows: By introducing a state evolution trend analysis mechanism, this invention can identify potentially risky state segments before obvious failures occur, enabling early warning and trend guidance for changes in key parameters during printing tasks, effectively reducing the risk of print quality degradation or equipment damage caused by sudden state changes. Simultaneously, this invention employs multimodal data-driven conditional modeling and state clustering, improving the predictive model's adaptability and generalization ability to complex operating conditions, making it particularly suitable for real-world industrial environments with frequent printing task switching and fluctuating equipment operating states. Through trend consistency judgment and state label update mechanisms, a state prediction sequence with evolutionary significance can be output, providing a decision-making basis for subsequent task scheduling and maintenance intervention. In summary, this invention can achieve dynamic evolution modeling and trend prediction of the printing equipment's operating state, thereby effectively improving the system's ability to identify and respond to potential anomalies, cold-start unstable states, and modal fluctuations. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart of a multimodal printer state prediction method;

[0026] Figure 2 This is a structural diagram of a multimodal printer state prediction system. Detailed Implementation

[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0030] As mentioned in the background section, existing printer status correction or control methods mostly focus on single-modal data modeling or local control optimization of unit hardware. They have not yet formed a complete framework for fusing multimodal data and establishing an evolution trend prediction model, especially lacking accurate status identification and prediction mechanisms under cold start and sudden modal fluctuation conditions. Therefore, a multimodal printer status prediction scheme is needed.

[0031] Figure 1 This is a flowchart of a multimodal printer state prediction method according to an embodiment of the present invention. Figure 1 As shown, the multimodal printer state prediction method includes the following steps:

[0032] S1: Obtain the multimodal data stream set, determine whether it is in the cold start phase, and output the phase label.

[0033] First, the multimodal data stream includes, but is not limited to, printhead temperature, feed rate, power consumption characteristics, and electronic control signals. In this embodiment of the invention, in order to determine whether the acquired multimodal data stream is in the cold start phase, a comprehensive judgment is made based on the fluctuation of data characteristics within the initial time window.

[0034] Ideally, the determination of the cold start phase includes the following steps:

[0035] 1) Construct a window for the first k time slices in the multimodal data stream set, extract the modal feature component set, and calculate the heteroscedasticity for each modal feature to form a feature heteroscedasticity vector, where k is a constant.

[0036] It should be noted that heteroscedasticity refers to the non-uniformity of fluctuations in feature values ​​across different time slices, effectively reflecting the instability level of features in the initial stage. Traditional state judgment often relies on basic statistical indicators such as mean and variance. However, this invention, by employing the heteroscedasticity index, can more sensitively capture the drastic fluctuations or local instability that may exist in the early stages of cold start, thereby improving the accuracy of cold start judgment. In specific operation, for the feature sequence of each modal channel, the local variance of each time slice within the window is calculated sequentially, and weighted according to the weight of the time slice position. Finally, the heteroscedasticity value under a single modality is output. The set of heteroscedasticity values ​​of all modalities constitutes the feature heteroscedasticity vector in the multimodal case.

[0037] 2) Calculate the volatility density within the current window based on the feature heteroscedasticity vector, and calculate the ratio of the average volatility density stored in the historical stable phase as a cold start discriminant factor.

[0038] Specifically, the fluctuation density is a weighted sum of the values ​​of each dimension of the feature heteroscedasticity vector, which can intuitively reflect the overall level of fluctuation of the current window features. The purpose of this step is to uniformly measure the fluctuation trend of the system state from the multimodal feature level, avoid the one-sided judgment caused by the anomaly of a single feature, and thus improve the consistency and reliability of state discrimination.

[0039] 3) If the ratio is greater than or equal to the density offset threshold, the output stage label is cold start; otherwise, the output stage label is stable stage.

[0040] Specifically, this invention calculates the ratio of the fluctuation density value obtained in the current window to the average fluctuation density stored in historical stable phases to form a cold start discrimination factor. If the ratio is greater than or equal to a preset density offset threshold, the current multimodal data stream is determined to be in the cold start phase; otherwise, it is determined to be in the stable phase. The density offset threshold is a preset parameter of the system, and its value is set according to the tolerable fluctuation range and the requirements for misjudgment risk control in the specific application scenario.

[0041] Of particular note is that, through the aforementioned cold start phase discrimination steps, this invention enables rapid and accurate segmentation of state stages at the initial stage of multimodal data acquisition, providing a phased basis for subsequent data processing, state prediction, and anomaly detection. Compared to traditional methods that rely solely on time periods for simple judgment, the method described in this invention can achieve phase segmentation at the operational data feature level based on dynamic comparison of characteristic fluctuations and historical fluctuation levels, exhibiting stronger adaptability and higher accuracy. Furthermore, the design of this cold start discrimination factor can be extended to the detection of other state transition nodes, such as equipment startup, load surges, or system resets, providing a generalized discrimination framework for state management.

[0042] S2: Perform conditional partitioning of the data stream set based on the stage label, and calculate the feature vector for each partition.

[0043] It's important to note that conventional data processing often ignores stage attributes, relying solely on time windows for slicing. This can easily lead to data from different operational states being mixed together, affecting the accuracy of feature extraction. This invention, however, uses stage labels as the slicing basis, ensuring that each slice corresponds only to data from the same operational state (e.g., cold start or stable phase), thus guaranteeing stage consistency in subsequent feature vector calculations. Through this operation, precise binding between data and operational state can be effectively achieved, making feature extraction more discriminative and interpretable. The specific operation is as follows:

[0044] A better approach to conditional partitioning includes the following steps:

[0045] A) Map the stage labels to the multimodal data stream set, and for each stage label, extract the data stream sequence generated in the corresponding stage.

[0046] The above operations are based on jointly indexing the timestamps and stage labels in the multimodal data stream set. For time segments labeled "cold start," the corresponding multimodal data stream sequence is extracted. Similarly, for time segments labeled "stable stage," the same data filtering is performed, avoiding the confusion of features from different stages. This ensures the independence and purity of cold start and stable operation features in subsequent calculations. This avoids the risk of the sliding window method used in conventional methods, which involves the window spanning two operating states. Instead, this invention strictly demarcates data using stage labels, ensuring that each data segment belongs to a single state, achieving high reliability and state purity in the data sharding process.

[0047] B) Based on the temporal order between data points, the data stream sequence is divided into several continuous data segments, each with a fixed period. Specifically, the fixed period can be preset according to the device operating characteristics and data sampling frequency, such as dividing it into units of 10 seconds, 30 seconds, or 1 minute. This embodiment of the invention does not impose a unique limitation, thereby ensuring that each data segment has sufficient length for statistical feature extraction, while avoiding excessive length that would cause state fluctuations to be smoothed out.

[0048] C) Within each segment, based on whether the printhead temperature exceeds the upper or lower fluctuation range of the average value of the corresponding stage, the data points within the segment are further divided into segments to generate a conditional segment set.

[0049] Specifically, the mean printhead temperature μ and standard deviation σ are first calculated from historical data within the current stage (such as cold start or stabilization stage), and a fluctuation range threshold (such as μ ± 1.5σ) is set. During operation, the data points within each time period are traversed to determine whether their printhead temperature falls within the set range. If it exceeds the range, it is determined to be a temperature fluctuation point, and based on this, the original data segment is divided into multiple conditional segments.

[0050] This operation differs from existing simple methods that divide data by fixed time or number of data points. It is highly adaptive and state-aware, accurately capturing the implicit correlation between printhead temperature fluctuations and device operating status. Through this conditional slicing, not only is the temporal continuity of the original data preserved, but it also enables effective response to dynamic changes in key device parameters and adjustment of data slices, thus providing more representative fragment data for subsequent feature vector calculations.

[0051] Furthermore, calculating the feature vector for each piece includes the following steps:

[0052] For each conditional segment, data standardization is first performed to eliminate the dimensional influence between different modalities and sampling scales. For each conditional segment, statistical indicators of all modalities in the segment are collected to construct its segment feature vector, including but not limited to mean and variance. The above multiple statistical features are concatenated into a feature vector of uniform length, which serves as the representative expression of the current conditional segment.

[0053] As can be seen, this invention can dynamically reflect the subtle fluctuations in equipment operation. For example, during the cold start phase, the printhead temperature fluctuates significantly, and conditional fragmentation generates more and shorter segments to meticulously depict the gradual temperature rise during the cold start process. In the stable phase, the temperature fluctuation range is smaller, and the system generates longer and more uniform conditional segments, thus reflecting the stability of equipment operation. Compared to the limitations of existing fragmentation methods that cannot detect changes in printhead state, this invention introduces printhead temperature as a key parameter for fragmentation, achieving a linkage adjustment between data fragmentation and the actual operating state of the equipment, thereby significantly improving the representational ability of data features and the model training effect.

[0054] S3: Construct a local polynomial regression model for each segment, and map the trend vector of key indicators in the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, mark the current segment as a potential abnormal segment.

[0055] S3.1: The construction of a local polynomial regression model includes:

[0056] S3.1.1: For any feature vector, retrieve the last T in the corresponding conditional slice. eThe invention presents a sequence of key indicators per unit time period. These key indicators include, but are not limited to, modal values ​​of printhead temperature and feed rate. It should be noted that, in selecting key indicators, this invention prioritizes modal values ​​that are sensitive to changes in the equipment's operating state, such as printhead temperature and feed rate. Printhead temperature reflects the dynamic trend of the equipment's heating system, while feed rate is directly related to the stability of the material supply system. These two types of indicators provide fundamental data support for predicting equipment state trends. Therefore, for each segment, the system automatically extracts the aforementioned key indicator value sequence at the end of the time period as the data basis for subsequent polynomial fitting models. This differs from existing technologies that use overall mean or static data for the entire segment for modeling, achieving rapid capture and processing of dynamic trends in local time periods.

[0057] S3.1.2: Introduce a weight sequence to the key indicator sequence, where each weight is dynamically calculated based on the modal variance and historical mean of that data point, so that high volatility points account for a higher proportion in the fitting, and construct a weighted vectorized trend sequence s. k .

[0058] The weight calculation logic is as follows: if a data point exhibits high volatility (i.e., large modal variance) within the current window, or if its value deviates significantly from the historical average, the system assigns that point a higher weight. Through this mechanism, this invention allows data points with high volatility or large deviations to dominate the fitting process, thereby effectively amplifying subtle anomalies and improving trend detection accuracy.

[0059] S3.1.3: Based on s k The key indicator sequence is constructed using a polynomial fitting modeling method of no more than order 3, and a local prediction expression is independently constructed for each modal indicator.

[0060] Specifically, based on the aforementioned weighted key indicator sequence, this invention employs a polynomial fitting method of order no more than 3 to independently construct a local prediction model for each modal indicator. The polynomial order is controlled within 3 to avoid overfitting while ensuring that the model has the ability to describe nonlinear trend changes. For example, for the printhead temperature indicator, the fitting model can be expressed as: T(t)=a0+a1t e +a2t e 2 +a3t e 3 Where T(t) is the prediction time t eThe temperature values ​​at each time point, a0 to a3, are polynomial coefficients obtained using the weighted least squares method. Similarly, an independent prediction expression is constructed for the feed rate index. In this process, based on the segmented features described by the feature vectors, adaptive fitting of the polynomial model parameters is achieved, enabling the model to accurately map to the local dynamic features of the current segment, thereby improving the targeting and accuracy of the prediction. Through the polynomial regression model, this invention can predict the changing trends of key indicators over the next 3 to 5 periods, providing a trend basis for subsequent abnormal segment identification.

[0061] After completing the trend vector prediction, the present invention further implements an abnormal segment confirmation mechanism, which determines whether there are potential anomalies in the current segment by analyzing the difference between the trend vectors.

[0062] S3.2: Identification of abnormal segments includes the following steps:

[0063] First, calculate the continuous difference Δ for the trend vector. k And construct a trend mutation factor group, where any trend mutation factor is |Δ j+1 -Δ j |; Combine all trend-changing factor groups in the historical segments of each stage to calculate the stability reference interval. in, and The mean and standard deviation of the mutation factors are represented by θ, and θ is the adjustment threshold coefficient. The mechanism determines whether any mutation factor in the trend mutation factor group exceeds the stability reference interval Λ. If it does, the current segment is marked as a deviation segment, and the trigger point and corresponding time point are recorded. This mechanism differs from existing methods that only detect single features such as mean deviation or volatility anomalies, achieving accurate capture of trend mutation features and effectively improving the accuracy and reliability of anomaly detection. If it is a deviation segment, the multimodal data set contained in the feature vector of that segment is extracted, the variation amplitude within the current window is calculated for each mode, and a modal fluctuation sequence is constructed. If the variation amplitude of at least two items in the modal fluctuation sequence is higher than the upper bound of the historical modal interval, the segment is confirmed as an anomalous segment. This dual verification mechanism—trend mutation detection and modal fluctuation verification—effectively avoids misjudgment caused by single-indicator anomalies, ensuring that the marked segments are true anomalous segments and possessing high detection robustness.

[0064] The variation amplitude is defined as the difference between the maximum and minimum values.

[0065] S4: Cluster and sort the set of potential abnormal segments in the time dimension, generate a state-level transition sequence through the transition association between segments, and output the state prediction result based on trend consistency.

[0066] S4.1: The generation of the state-level transition sequence includes the following steps:

[0067] The set of abnormal segments is used to construct a monotonically increasing indexed linked list of segments based on timestamp information, and the time interval parameter γ between segments is embedded in the indexed linked list structure. t This parameter is used to determine the temporal correlation and attribution relationship between segments and to establish a temporal sequence constraint structure. In this way, a segment organization structure with temporal sequence constraints is formed, laying the foundation for clustering in the time dimension.

[0068] Furthermore, adjacent potentially anomalous fragments are placed in γ t Within the range, segments are grouped into clusters at the same time, denoted as TQ. The same rule is applied to all consecutive segments that meet the conditions, and they are uniformly grouped into one TQ. Each TQ is then assigned a segment sequence label, forming a sequence of segment clusters aggregated along the time dimension.

[0069] Furthermore, after constructing the TQ sequence, the process proceeds to the state transition path determination stage. The core of this stage lies in determining whether a valid state transition path exists between different TQs based on the consistency ratio of the segments within each TQ in the characteristic trend direction. Specifically, this is based on the consistency ratio η of the segments within each TQ in the characteristic trend direction. h Construct a cross-cluster transfer criterion, the criterion satisfying η h When the consistency threshold is greater than a preset directional threshold, a state transition path from TQ to TQ+1 is established. Based on the establishment of each state transition path, a state-level transition sequence is generated, and each pair of consecutive cluster associations in the transition sequence is marked as a first-level state transition pair. The consistency ratio η... h Defined as the ratio of the number of pairs of segments with consistent direction to the total number of segments. It can be seen that this sequence fully describes the dynamic evolutionary link formed by the combination of temporal aggregation, feature trend correlation, and state transition relationships of potential anomalous segments.

[0070] S4.2: Outputting state prediction results based on trend consistency includes the following steps:

[0071] For each pair of consecutive segment clusters, compare whether the overall trend direction of the corresponding segments in the original sequence remains continuous. The specific calculation method is as follows: calculate the mean of the trend vectors for TQ and TQ+1 respectively, calculate the angle between the two, and if the angle is less than a set threshold (such as 10 degrees), then the trend consistency is considered to be valid. When trend consistency is established, it is assumed that the state labeled in the previous segment cluster remains unchanged in the subsequent segment cluster. In this case, the state label of the previous segment cluster is continued and assigned to the subsequent segment cluster. The two are regarded as a continuous evolution process of the state, which is represented in the form of smooth state continuation in the state-level transition sequence. If the trend is judged to be inconsistent, it is regarded as a potential state transition. In this case, the state label of the previous segment cluster is not directly continued. The current segment cluster is independently labeled, and the position of the segment cluster is marked as a state change node in the state-level transition sequence. After completing the trend consistency judgment and state label processing for all continuous segment clusters, the final state label of each segment cluster is rearranged according to the original time order, and a state prediction result sequence covering all segment clusters is output. This sequence not only preserves the transition relationship in the time dimension, but also clearly reveals the continuation and transition nodes of the state, serving as the final trend-guided state prediction output.

[0072] S5: Based on the comparison between the state prediction results and the subsequent real state labels, adjust the weights of each regression model accordingly.

[0073] First, using the state prediction result sequence output in step S4 as input, a step-by-step comparison with the subsequent true state labels is performed. Specifically, the predicted state label for each segment cluster in the state prediction result sequence is extracted, and the corresponding true state label in time is also extracted. This process is repeated for all segment clusters to determine the state label error between them.

[0074] For example, if the state label is categorical, a classification difference function can be used. For instance, if the predicted state label equals the true state label, the state bias is 1; otherwise, it is 0. If the state label is numerical, Euclidean distance or absolute difference can be used for quantification.

[0075] By comparing the state labels of all segment clusters in the sequence, a prediction error vector is constructed. This error vector fully reflects the local deviation distribution characteristics of the prediction results across different segment clusters. To further quantify the uniformity of the error distribution, an error distribution uniformity index G is calculated, defined as the ratio of the standard deviation to the mean of the error vector. This index characterizes the fluctuation range of the error in the sequence. A large G indicates an uneven local error distribution, with some segment clusters exhibiting excessively high errors; conversely, a small G indicates a relatively smooth and consistent error distribution. When the error distribution uniformity index G exceeds a preset range, the model weight adjustment mechanism is activated.

[0076] Weight adjustment includes dynamically adjusting the local weight factors of each model based on the error vector. The adjustment strategy is as follows: first, analyze the peak position of the error vector to identify clusters of high-value segments within the error set; then, based on their historical performance, appropriately reduce the weight of models with higher errors while increasing the weight of models with lower errors. The weight adjustment uses a scaling method to ensure that the model combination automatically strengthens its fitting ability to error-sensitive regions in subsequent predictions.

[0077] Furthermore, this embodiment also provides a multimodal printer status prediction system, including: a cold start discrimination module 100, used to acquire a multimodal data stream set, determine whether it is in the cold start stage, and output a stage label;

[0078] The conditional sharding module 200 is used to conditionally shard the data stream set according to the stage label and calculate the feature vector for each shard.

[0079] The anomaly labeling module 300 is used to construct a local multinomial regression model for each segment, and to map the trend vector of key indicators in the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, the current segment is labeled as a potential anomaly segment.

[0080] The state transition module 400 is used to perform time-dimensional clustering and sorting of the set of potential abnormal segments, generate a state-level transition sequence through the transition association between segments, and output the state prediction result based on trend consistency.

[0081] The model adjustment module 500 is used to adjust the weights of each regression model based on the comparison between the state prediction results and the subsequent real state labels.

[0082] This embodiment also provides a computer device suitable for the multimodal printer state prediction method, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multimodal printer state prediction method proposed in the above embodiment.

[0083] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0084] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the method for predicting the state of a multimodal printer as proposed in the above embodiments.

[0085] In summary, by introducing a state evolution trend analysis mechanism, this invention can identify potentially risky state segments before obvious failures occur, enabling early warning and trend guidance for changes in key parameters during printing tasks, effectively reducing the risk of print quality degradation or equipment damage caused by sudden state changes. Simultaneously, this invention employs multimodal data-driven conditional modeling and state clustering, improving the predictive model's adaptability and generalization ability to complex operating conditions, making it particularly suitable for real-world industrial environments with frequent printing task switching and fluctuating equipment operating states. Through trend consistency judgment and state label update mechanisms, a state prediction sequence with evolutionary significance can be output, providing a decision-making basis for subsequent task scheduling and maintenance intervention. In conclusion, this invention enables dynamic evolution modeling and trend prediction of printing equipment operating states, thereby effectively improving the system's ability to identify and respond to potential anomalies, cold-start instability states, and modal fluctuations.

[0086] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the state of a multimodal printer, characterized in that: include: Acquire a multimodal data stream set, determine whether it is in the cold start phase, and output the phase label; The data stream set is conditionally partitioned based on stage labels, and feature vectors are calculated for each partition. Construct a local polynomial regression model for each segment, and map the trend vector of key indicators in the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, mark the current segment as a potential abnormal segment. The potential abnormal fragment set is clustered and sorted in the time dimension, and a state-level transition sequence is generated through the transition association between fragments. The state prediction result is output based on trend consistency. Based on the comparison between the state prediction results and the subsequent real state labels, the weights of each regression model are adjusted accordingly. The generation of the state-level transition sequence includes: constructing an indexed linked list of monotonically increasing segments from the set of anomalous segments according to timestamp information, and embedding the inter-segment time interval parameter γ in the indexed linked list structure. t Establish a temporal sequence constraint structure; group adjacent potential anomalous segments in γ t Fragments within the same time frame are grouped into clusters, denoted as TQ, and each TQ is assigned a fragment sequence label, forming a fragment cluster sequence aggregated along the time dimension; based on the consistency ratio η of fragments within each TQ in the characteristic trend direction... h Construct a cross-cluster transfer criterion, the criterion satisfying η h When the value exceeds the preset directional consistency threshold, a state transition path from TQ to TQ+1 is established; Based on the state transition path, a state-level transition sequence is generated, and each pair of consecutive cluster associations in the transition sequence is marked as a first-level state transition pair. For each pair of consecutive segment clusters, compare whether the overall trend direction of the corresponding segments in the original sequence remains continuous; after completing the trend consistency judgment and state label processing for all consecutive segment clusters, rearrange the final state labels of each segment cluster according to the original time order, and output a state prediction result sequence covering all segment clusters as the final trend-guided state prediction output.

2. The multimodal printer state prediction method as described in claim 1, characterized in that: The determinations during the cold start phase include: For the first k time slices in the multimodal data stream set, construct a window, extract the modal feature component set, and calculate the heteroscedasticity for each modal feature to form a feature heteroscedasticity vector, where k is a constant; The volatility density within the current window is calculated based on the feature heteroscedasticity vector, and the ratio of the average volatility density stored in the historical stable phase is calculated as a cold start discriminant factor. If the ratio is greater than or equal to the density offset threshold, the output stage label is cold start; otherwise, the output stage label is stable stage.

3. The multimodal printer state prediction method as described in claim 1, characterized in that: The conditional partitioning includes: The stage labels are mapped to the multimodal data stream set, and for each stage label, the data stream sequence generated in the corresponding stage is extracted; Based on the time order between data points, the data stream sequence is divided into several continuous data segments, each with a fixed period of length; Within each segment, the data points within the segment are further divided based on whether the printhead temperature exceeds the upper or lower fluctuation range of the average value of the corresponding stage, generating a conditional segment set.

4. The multimodal printer state prediction method as described in claim 1, characterized in that: The construction of the local polynomial regression model includes: For any feature vector, retrieve the last T in the corresponding conditional slice. e The sequence of key indicators per unit time; among which, key indicators include, but are not limited to, modal values ​​of printhead temperature and feed rate; A weight sequence is introduced onto the key indicator sequence, where each weight is dynamically calculated based on the modal variance and historical mean of the key indicator sequence data points, so that high volatility points account for a higher proportion in the fitting, and a weighted vectorized trend sequence s is constructed. k ; Based on s k The key indicator sequence is constructed using a polynomial fitting modeling method of no more than order 3, and a local prediction expression is independently constructed for each modal indicator.

5. The multimodal printer state prediction method as described in claim 4, characterized in that: The confirmation of the abnormal segment includes: Calculate the continuous difference Δ for the trend vector k And construct a trend mutation factor group, where any trend mutation factor is |Δ j+1 -Δ j |; By combining all trend-changing factor groups in the historical segments of the stage, the stability reference interval Λ = [μ φ -θ×σ φ ,μ φ +θ×σ φ ], where μ φ and σ φ θ represents the mean and standard deviation of the mutation factor, respectively, and θ is the adjustment threshold coefficient; determine whether any mutation factor in the trend mutation factor group exceeds the stability reference interval Λ. If it does, mark the current partition as a deviation partition and record the trigger point and the corresponding time point. If it is a deviated segment, the multimodal data group contained in the feature vector of the segment is extracted, the variation amplitude within the current window is calculated for each mode, and a modal fluctuation sequence is constructed; if the variation amplitude of at least two items in the modal fluctuation sequence is higher than the upper bound of the historical modal interval, the segment is confirmed as an abnormal segment.

6. The multimodal printer state prediction method as described in claim 1, characterized in that: The trend-consistency-based output state prediction results include: When trend consistency is established, it is assumed that the state marked by the previous segment cluster remains unchanged in the next segment cluster. In this case, the state label of the previous segment cluster is continued to be assigned to the next segment cluster, which is regarded as a continuous evolution process of the state. If an inconsistency is found in the trend, it is considered a potential state transition. In this case, the state label of the previous fragment cluster is not directly continued. Instead, the current fragment cluster is independently labeled, and its position is marked as a state change node in the state transition sequence.

7. A multimodal printer status prediction system, based on the multimodal printer status prediction method according to any one of claims 1 to 6, characterized in that: Also includes: The cold start discrimination module is used to acquire a multimodal data stream set, determine whether it is in the cold start phase, and output a phase label; The conditional sharding module is used to conditionally shard the data stream collection based on the stage label and calculate the feature vector for each shard. The anomaly labeling module is used to construct a local multinomial regression model for each segment, and to map the trend vector of key indicators in the next 3 to 5 periods through each feature vector; if the conditions of trend change and modal fluctuation are met, the current segment is labeled as a potential anomaly segment. The state transition module is used to cluster and sort the set of potential abnormal segments in the time dimension, generate a state-level transition sequence through the transition association between segments, and output the state prediction result based on trend consistency. The model adjustment module is used to adjust the weights of each regression model based on the comparison between the state prediction results and the subsequent real state labels.

8. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the multimodal printer state prediction method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the multimodal printer state prediction method according to any one of claims 1 to 6.

Images