A real-time acquisition and intelligent analysis method for energy consumption data of an injection molding production line

Abstract

The present application relates to the technical field of electric digital data processing, in particular to a kind of real-time acquisition and intelligent analysis method of injection molding production line energy consumption data, comprising: obtaining the multidimensional energy consumption parameter data of current time node and historical multiple time nodes on injection molding production line, multidimensional energy consumption parameter data of any time node is regarded as a data point;The local outlier factor score of the data point corresponding to the current data point is calculated using improved local outlier factor algorithm, if the local outlier factor score is greater than the preset abnormal threshold, it is determined that there is an abnormality in the current time node.The present application solves the problem of high complexity of existing algorithm.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology. More specifically, this invention relates to a method for real-time acquisition and intelligent analysis of energy consumption data in injection molding production lines. Background Technology

[0002] With the continuous development of the manufacturing industry towards digitalization and intelligence, injection molding production lines, as typical high-energy-consuming discrete manufacturing systems, involve multiple stages in their operation, including heating, melting, injection, pressure holding, and cooling. The energy consumption characteristics of each stage differ significantly and exhibit obvious temporal fluctuations. In actual production, injection molding equipment typically operates continuously for extended periods, with frequent switching between different operating conditions. Furthermore, it is affected by a combination of factors, including raw material characteristics, mold structure, ambient temperature, and equipment aging, resulting in energy consumption data exhibiting multi-dimensional, strongly coupled, and non-stationary changes. Therefore, real-time collection and intelligent analysis of energy consumption data from injection molding production lines are crucial for achieving refined energy management, reducing production costs, and improving equipment operational stability.

[0003] Currently, the analysis of industrial energy consumption data mostly adopts technical approaches based on statistical analysis or traditional anomaly detection methods. For example, it uses fixed threshold judgment, simple time series analysis, or standard local outlier detection methods to identify abnormal energy consumption behavior. However, in the typical complex operating scenario of an injection molding production line, existing methods often fail to achieve ideal results. On the one hand, traditional methods typically rely only on data features within a single moment or a limited time window, lacking in-depth modeling of historical trends. This makes it difficult to accurately depict the dynamic evolution of energy consumption data over time, easily misjudging normal operating fluctuations as anomalies, or reacting slowly to gradual abnormal changes. On the other hand, conventional local outlier analysis methods are mostly based on the assumption of a fixed neighborhood structure and uniform scale, failing to fully consider the unevenness of different data points in spatial distribution density, temporal correlation, and differences in operating states. This results in poor stability of anomaly detection results in complex environments where high and low loads alternate.

[0004] Existing methods typically employ static calculation mechanisms, performing uniform outlier calculations on all data points without differentiating the data based on its sensitivity or impact in the current state. This not only increases unnecessary computational burden but also makes the algorithm lack targeted response capabilities when facing localized drastic fluctuations or abnormal propagation. Consequently, existing algorithms suffer from high computational complexity, making it difficult to meet the needs of real-time analysis and causing system response delays. Summary of the Invention

[0005] To address the issue of high computational complexity in existing algorithms mentioned in the background section, this invention provides the following solution.

[0006] In a first aspect, the present invention provides a method for real-time acquisition and intelligent analysis of energy consumption data of an injection molding production line, comprising: acquiring multidimensional energy consumption parameter data of the injection molding production line at the current time node and multiple historical time nodes, and taking the multidimensional energy consumption parameter data of any time node as a data point; calculating the local outlier factor score of the data point corresponding to the current data point using an improved local outlier factor algorithm, and determining that there is an anomaly at the current time node if the local outlier factor score is greater than a preset anomaly threshold; wherein, the improved local outlier factor algorithm specifically comprises: if the current data point corresponds to If the vulnerability index of a target data point within its neighborhood is greater than a set threshold, the local outlier score is recalculated; if the vulnerability index is less than or equal to the set threshold, the local outlier score of the data point preceding the target data point is used as the local outlier score of the target data point. The target data point is... Any data point within the neighborhood; the vulnerability index is positively correlated with the radius of influence of the disturbance, and negatively correlated with the Euclidean distance between the target data point and the current data point in multidimensional space and the corresponding time difference; the radius of influence of the disturbance is positively correlated with the degree of anomaly of the target data point.

[0007] The aforementioned technical solution introduces an adaptive adjustment mechanism based on the strength of data influence and spatiotemporal correlation during energy consumption anomaly detection. This eliminates the need for a uniform approach in calculating outlier levels, allowing for dynamic adjustment based on the actual impact of neighboring data on the current state. Updates are made promptly when key fluctuations or potential anomalies have a strong impact, ensuring the sensitivity of detection results to anomaly changes. Simultaneously, existing results are directly inherited when data is relatively stable or the impact is weak, effectively reducing redundant calculations and improving overall processing efficiency. Furthermore, by incorporating spatial proximity and time intervals between data points as constraints, the analysis results focus more on data highly correlated with the current state, avoiding interference from distant or lagging data. Combined with characterization of the magnitude of anomaly changes, adaptive adjustment of the anomaly's impact range is achieved. Overall, this significantly improves the accuracy and stability of anomaly identification, reduces the probability of false positives and false negatives, and balances real-time performance with computational efficiency, resulting in stronger adaptability and reliability in complex operating conditions and multi-dimensional energy consumption coupling scenarios.

[0008] Furthermore, target data points Vulnerability Index for: , For target data points The radius of influence of the disturbance. For target data points Compared with the current data point Distance in multidimensional space , Target data points Current data point The corresponding time, Set the preset loop time.

[0009] The aforementioned technical solution couples the range of impact of data disturbances with their proximity in the feature space and their temporal correlation, enabling the evaluation results to simultaneously reflect the propagation capability of local fluctuations and the actual intensity of their impact on the current state. Data that are spatially closer and temporally more adjacent are given higher weights, thereby strengthening the ability to identify key impact data, while data that are far away or have a large time interval are effectively weakened to avoid interfering with the judgment results. This allows for a more accurate characterization of the sources of change that have a substantial impact on the current state, improving the targeting and sensitivity of anomaly identification, while enhancing the stability and anti-interference capability of the overall method under complex working conditions, and providing a more reliable basis for subsequent analysis.

[0010] Furthermore, target data points Disturbance radius for: , For the current data point All data points in the neighborhood The mean of the values, For the natural constant An exponential function with base 0. For target data points The degree of abnormality, , The corresponding data points are respectively The mean and standard deviation of the anomaly level of all data points within the neighborhood.

[0011] The aforementioned technical solution introduces a nonlinear adjustment for the relative deviation of anomalies at the local neighborhood scale. This allows the range of influence to adaptively expand and contract with changes in the data relative to the overall surrounding level. When a data point exhibits a significant deviation, its range of influence is rapidly amplified, highlighting its potential anomaly impact. Conversely, it maintains relative convergence when approaching the overall distribution level, avoiding interference from ineffective expansion. Simultaneously, normalization is performed based on the dispersion of the distribution within the neighborhood, ensuring consistency in evaluation results under different densities and fluctuation environments. This enhances the sensitivity to local anomalies, improves the differentiation of anomaly intensity differences, and maintains good stability and adaptability under complex data distribution conditions, thereby improving the accuracy and reliability of the overall analysis results.

[0012] Furthermore, target data points abnormality for: , , Target data points Target data points The power at the corresponding time point, , Target data points Target data points The corresponding time, No-load power, It is the hyperbolic tangent function. The set time window length.

[0013] The aforementioned technical solution comprehensively characterizes the rate of change between the current state and historical states, and introduces a nonlinear adjustment mechanism to adaptively scale changes under different operating intensities. This allows the evaluation results to sensitively reflect drastic fluctuations in a short period of time while also reasonably constraining the magnitude of changes in conjunction with the overall operating level, thus maintaining consistent discrimination capability under different load conditions. Simultaneously, by accumulating and processing change information from multiple moments, it effectively enhances the comprehensive perception capability of both persistent and sudden anomalies, avoiding the one-sidedness caused by relying solely on a single moment or a single change feature. This significantly improves the accuracy and stability of identifying abnormal behavior, reduces the probability of misjudging normal fluctuations, and enhances the adaptability and reliability of the method in complex working conditions and variable operating environments.

[0014] Furthermore, the multidimensional energy consumption parameter data includes active power, reactive power, and current.

[0015] Furthermore, it also includes denoising and standardizing the multidimensional energy consumption parameter data.

[0016] Furthermore, the denoising employs an adaptive Kalman filter algorithm.

[0017] Furthermore, the standardization process employs Z-score standardization.

[0018] Furthermore, the distance is a Euclidean distance.

[0019] Furthermore, it also includes issuing an early warning when the scores of local outliers at multiple consecutive time points all exceed a preset anomaly threshold.

[0020] The aforementioned technical solution aggregates and judges anomaly assessment results from continuous time points, transforming single anomaly signals into a mechanism for identifying continuous anomaly trends. This effectively distinguishes between instantaneous fluctuations and genuine abnormal states, enabling the system to trigger timely warnings when abnormal behavior persists, providing early alerts of potential risks, while avoiding false alarms due to short-term fluctuations, thus improving the overall reliability and stability of monitoring. Through this continuous judgment, the method enhances the ability to perceive anomaly development trends, providing more valuable information for timely intervention and decision-making, and improving the responsiveness and practicality of the entire energy consumption data monitoring system.

[0021] The beneficial effects of this invention are as follows: This invention improves the local outlier factor algorithm, fully utilizing the spatial distribution characteristics and historical evolution patterns of data within the neighborhood. It also comprehensively incorporates the degree of data anomaly, the range of disturbance impact, and temporal correlation into the evaluation, adaptively calculating the sensitivity and potential influence of each data point on the current state. This enables anomaly identification to dynamically respond to multi-dimensional coupled fluctuations under complex operating conditions. By introducing the anomaly propagation radius and vulnerability index, this invention effectively distinguishes between data points with actual impact on the current state and relatively stable or weakly influential data. This maintains high sensitivity while suppressing misjudgments, enhancing the stability and robustness of the system under different loads and operating stages. In summary, this invention achieves high-precision, real-time, and dynamic monitoring of energy consumption data under complex and variable injection molding production conditions, enhancing the accuracy and reliability of anomaly identification. It also provides forward-looking and actionable decision-making support, offering solid technical support for energy consumption optimization, energy saving, and safe and stable operation of the production line, thereby significantly improving the intelligence level and practical value of the entire monitoring and analysis system. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to an embodiment of the present invention; Figure 2 This is a radar chart illustrating the comprehensive performance comparison before and after algorithm improvement of a real-time acquisition and intelligent analysis method for energy consumption data of an injection molding production line according to an embodiment of the present invention. Detailed Implementation

[0023] An embodiment of a method for real-time acquisition and intelligent analysis of energy consumption data in injection molding production lines.

[0024] like Figure 1 As shown in the flowchart, an embodiment of the present invention provides a method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line, comprising the following steps: S1: Obtain multi-dimensional energy consumption parameter data of the injection molding production line at the current time node and multiple historical time nodes, and take the multi-dimensional energy consumption parameter data of any time node as a data point.

[0025] In a preferred embodiment, the multidimensional energy consumption parameter data includes at least key electrical parameters such as active power, reactive power, and current, which can comprehensively reflect the operating status and energy utilization of the equipment. Through the coordinated collection and fusion characterization of multidimensional parameters, it helps to more accurately depict the energy consumption characteristics of the injection molding equipment under different operating conditions, and improve the comprehensiveness and reliability of subsequent analysis.

[0026] Furthermore, to reduce the uncertainty caused by sensor noise, environmental interference, and instantaneous fluctuations in the raw collected data, the multidimensional energy consumption parameter data is preprocessed. The denoising process employs an adaptive Kalman filter algorithm, which dynamically adjusts the weights of state noise and observation noise during the filtering process to achieve adaptive matching of data noise levels under different operating conditions. This effectively suppresses random noise and abnormal fluctuations while preserving the true energy consumption trend, significantly improving the smoothness and reliability of the data. Compared to fixed-parameter filtering methods, this approach is better suited to the complex and ever-changing operating environment during injection molding production, enhancing the robustness of data processing.

[0027] After denoising, the multidimensional energy consumption parameter data is further standardized, preferably using the Z-score standardization method. This maps energy consumption parameters of different dimensions and scales to a unified standard distribution space, ensuring comparability and consistency in numerical range across dimensions and avoiding model bias issues caused by differences in dimensions. This standardization process not only improves the convergence speed in subsequent data analysis and modeling but also enhances the model's ability to learn balanced energy consumption characteristics, improving the accuracy of anomaly detection and energy efficiency assessment.

[0028] S2: Calculate the local outlier score of the data point corresponding to the current data point using the improved local outlier algorithm.

[0029] like Figure 2 As shown in the radar chart, this invention provides a comparison of the comprehensive performance of a real-time data acquisition and intelligent analysis method for energy consumption in an injection molding production line before and after algorithm improvement.

[0030] In a preferred embodiment, the improved local outlier factor algorithm specifically involves: if the current data point corresponds to If the vulnerability index of a target data point within its neighborhood is greater than a set threshold, the local outlier score is recalculated; if the vulnerability index is less than or equal to the set threshold, the local outlier score of the data point preceding the target data point is used as the local outlier score of the target data point. The target data point is... Any data point within the neighborhood; By introducing a data stability discrimination mechanism into the traditional local outlier detection process, and using the strength of the influence of each data point in the neighborhood on the current state as the trigger condition, this method adaptively decides whether to recalculate the outlier degree. The spatial and temporal sensitivity of data points is used as the control criterion. When relevant data has a strong influence on the current state, recalculation is performed to ensure that the judgment result reflects the latest changes in a timely manner. When relevant data is relatively stable or has a weak influence, existing results are directly inherited to avoid redundant calculations, thus achieving a balance between dynamic updates and result continuation. This significantly reduces unnecessary computational overhead, improves algorithm efficiency, and ensures high response speed and recognition accuracy even when key fluctuations or anomalies occur. It avoids misjudgments caused by frequent updates or excessive lag, and overall improves the real-time performance, stability, and reliability of the method under complex conditions.

[0031] Target data points Vulnerability Index for: , For target data points The radius of influence of the disturbance. For target data points Compared with the current data point Distance in multidimensional space , Target data points Current data point The corresponding time, The preset cycle time is used, and the distance is a Euclidean distance.

[0032] By using the existing perturbation range as a foundation, and incorporating the proximity of current and historical data in a multi-dimensional feature space as well as the temporal interval relationship, a comprehensive evaluation mechanism that balances spatial similarity and temporal correlation is constructed. Spatial distance is used to characterize the similarity between different data states, while time interval normalization weakens the influence of older data on the current state. This ensures that the evaluation results reflect the intensity of local perturbations on similar states while avoiding interference from distant historical data. It enables more accurate identification of data points with actual impact on the current operating state, improves sensitivity to key anomaly propagation paths, and effectively suppresses interference from spatially distant or temporally weakly correlated data. This enhances the relevance, stability, and reliability of the overall evaluation results, providing a more valuable basis for subsequent anomaly detection and energy consumption optimization.

[0033] Target data points Disturbance radius for: , For the current data point All data points in the neighborhood The mean of the values, For the natural constant An exponential function with base 0. For target data points The degree of abnormality, , The corresponding data points are respectively The mean and standard deviation of the anomaly level of all data points within the neighborhood.

[0034] By Using spatial distribution characteristics within the neighborhood as a basic scale, the density of local data distribution is characterized. Then, relative deviation information of the degree of anomaly is introduced on this basis, so that the final influence range can not only reflect local structural characteristics, but also dynamically respond to the degree of anomaly of the current data relative to the overall level of the neighborhood. The logic is essentially to fuse spatial neighborhood information with statistical distribution information and use nonlinear mapping to enhance the ability to express differences. When the data shows obvious anomalies, its influence range can be significantly amplified, thereby improving the sensitivity and discrimination ability of outliers. When the data is close to the normal distribution, it automatically converges to a more stable range, effectively avoiding misjudgment problems caused by overexpansion. At the same time, it takes into account the adaptability under different density regions, improving the stability, robustness and ability to characterize the characteristics of anomaly propagation in complex data environments.

[0035] Target data points abnormality for: , , Target data points Target data points The power at the corresponding time point, , Target data points Target data points The corresponding time, No-load power, It is the hyperbolic tangent function. The set time window length.

[0036] By differentially characterizing the power change rate between the current moment and multiple historical moments, and comprehensively accumulating these change rates, an anomaly assessment mechanism that reflects the dynamic fluctuation characteristics of energy consumption is constructed. On this basis, a nonlinear modulation function related to the equipment's baseline operating state is further introduced, so that the anomaly assessment results can not only reflect the drasticness of short-term changes, but also adaptively adjust in combination with the overall operating load level. It can effectively distinguish between stable fluctuations under normal operating conditions and abrupt changes under abnormal operating conditions, avoiding misjudgment problems caused by a single amplitude or a single rate of change. At the same time, it has good adaptability in different operating ranges of low load and high load, improving the sensitivity and accuracy of anomaly identification, and enhancing the stability and reliability of the overall method under complex operating conditions.

[0037] S3: If the score of the local outlier factor is greater than the preset anomaly threshold, then it is determined that there is an anomaly at the current time point.

[0038] In a preferred embodiment, when the local outlier factor score calculated at any time point exceeds a preset anomaly threshold, it is determined that there is an abnormal state at that time point, thereby enabling timely identification of possible instantaneous abnormal behaviors during the production process and ensuring rapid response capability to abnormal energy consumption fluctuations.

[0039] Furthermore, in addition to identifying individual outlier data points, a continuous analysis mechanism is introduced. When the scores of local outliers at multiple consecutive time points exceed a preset anomaly threshold, the system triggers an early warning signal, indicating potential persistent or systemic anomaly risks. This continuous assessment effectively distinguishes between occasional, instantaneous fluctuations and genuine abnormal trends, reducing false alarms caused by short-term disturbances or normal equipment fluctuations, and improving the accuracy and stability of anomaly identification. Simultaneously, the continuous early warning mechanism enhances the ability to perceive the cumulative effects of anomalies, enabling the system to monitor the anomaly evolution process in real time, providing production managers with forward-looking risk warnings and scientific decision-making basis.

[0040] The present invention employs an improved local outlier factor algorithm, organically combining the distribution characteristics of data points in their spatial neighborhood, their multidimensional distance from the current state, and the time interval. It introduces an adaptive adjustment mechanism for the vulnerability index and the radius of influence of disturbances, ensuring that the degree of anomaly fully reflects the propagation capability of local fluctuations and the actual intensity of their impact on the current state. This approach can sensitively capture drastic fluctuations within a short period while suppressing normal fluctuations, avoiding misjudgments. In the analysis of continuous time nodes, the solution aggregates and judges the continuous trend of local outlier factor scores, transforming scattered instantaneous abnormal signals into a continuous abnormal trend identification mechanism. This effectively distinguishes between occasional fluctuations and genuine potential risks, enabling timely early warnings when abnormal behavior accumulates, providing early alerts of potential energy consumption anomalies in the production line, and reducing the possibility of false alarms and missed alarms. Overall, this solution achieves high-precision, dynamic, continuous, and adaptive analysis of energy consumption data from injection molding production lines. It not only significantly improves the sensitivity and stability of anomaly identification but also optimizes the system's real-time response capability and computational efficiency. This provides a reliable and scientific basis for energy consumption management, risk warning, and intelligent optimization of the production line, enhancing the safety, energy efficiency, and intelligence level of the production process, and meeting the actual needs of high-precision energy consumption monitoring and intelligent control under complex working conditions.

[0041] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0042] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for real-time acquisition and intelligent analysis of energy consumption data in injection molding production lines, characterized in that, include: Acquire multidimensional energy consumption parameter data of the injection molding production line at the current time node and multiple historical time nodes, and take the multidimensional energy consumption parameter data of any time node as a data point; The local outlier score of the data point corresponding to the current data point is calculated using an improved local outlier factor algorithm. If the local outlier score is greater than a preset anomaly threshold, it is determined that there is an anomaly at the current time point. Specifically, the improved local outlier factor algorithm is as follows: if the current data point corresponds to... If the vulnerability index of a target data point within its neighborhood is greater than a set threshold, the local outlier score is recalculated; if the vulnerability index is less than or equal to the set threshold, the local outlier score of the data point preceding the target data point is used as the local outlier score of the target data point. The target data point is... Any data point within the neighborhood; The vulnerability index is positively correlated with the radius of influence of the disturbance, and negatively correlated with the Euclidean distance between the target data point and the current data point in multidimensional space and the corresponding time difference; the radius of influence of the disturbance is positively correlated with the degree of anomaly of the target data point. Target data points Vulnerability Index for: , For target data points The radius of influence of the disturbance. For target data points Compared with the current data point Distance in multidimensional space , Target data points Current data point The corresponding time, Preset loop time; Target data points Disturbance radius for: , For the current data point All data points in the neighborhood The mean of the values, For the natural constant An exponential function with base 0. For target data points The degree of abnormality, , The corresponding data points are respectively The mean and standard deviation of the anomaly level of all data points within the neighborhood; Target data points abnormality for: , , Target data points Target data points The power at the corresponding time point, , Target data points Target data points The corresponding time, No-load power, It is the hyperbolic tangent function. The set time window length.

2. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 1, characterized in that, The multidimensional energy consumption parameter data includes active power, reactive power, and current.

3. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 1, characterized in that, It also includes denoising and standardizing the multidimensional energy consumption parameter data.

4. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 3, characterized in that, The noise reduction employs an adaptive Kalman filter algorithm.

5. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 3, characterized in that, The standardization process uses Z-score standardization.

6. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 1, characterized in that, The distance is a Euclidean distance.

7. The method for real-time acquisition and intelligent analysis of energy consumption data in an injection molding production line according to claim 1, characterized in that, Also includes: An alert is issued when the scores of local outliers at multiple consecutive time points exceed a preset anomaly threshold.

Images