Oil painting storage environment risk assessment method

Abstract

The invention discloses an oil painting storage environment risk assessment method, which comprises the following steps of: 1, acquiring environment data, and constructing a time sequence; 2, preprocessing the environment data to obtain environment time sequence data; carrying out slicing processing on the environment time sequence data, and constructing a feature sample used for time sequence prediction model input; 3, constructing an improved time sequence prediction model, predicting the environmental parameters of the oil painting based on the improved time sequence prediction model, and outputting a prediction result; and 4, based on the prediction result output in the step 3, performing modeling analysis on the long-term influence possibly generated by the predicted environmental condition on the oil painting material, and realizing the early recognition and quantitative evaluation of the oil painting storage environmental risk. According to the invention, through trend prediction and comprehensive evaluation of the environmental parameters, the environmental management is converted from passive response to early warning, and corresponding measures can be taken before the environmental abnormality has substantial influence on the oil painting.

Description

Technical Field

[0001] This invention relates to the fields of cultural relic environmental management and data processing technology, specifically to a method for risk assessment of the preservation environment of oil paintings. Background Technology

[0002] Oil paintings are composed of various materials such as canvas, base coat, pigment layer and protective coating. Under the long-term effects of environmental factors such as temperature, relative humidity, light, ultraviolet radiation and air pollutants, these materials will gradually undergo irreversible deterioration such as aging, fading, cracking or pollution absorption.

[0003] Currently, most museum oil painting preservation environment monitoring methods employ real-time monitoring based on fixed thresholds. This involves deploying environmental sensors such as temperature, humidity, light, and ultraviolet radiation in the exhibition space or storage room to continuously collect environmental parameters and compare the real-time values ​​with pre-set safety thresholds. When an environmental parameter exceeds the threshold, the system triggers an alarm, prompting management to take appropriate control measures. However, this threshold-based environmental monitoring method essentially only focuses on the instantaneous state of environmental parameters at a single moment. Its risk assessment logic is based on a binary judgment of "whether the threshold is exceeded," making it difficult to reflect the continuous changes in environmental conditions over time. When environmental parameters remain close to the threshold for an extended period but do not actually exceed it, this method typically does not trigger any risk warnings. However, the oil painting materials may continue to be adversely affected in this "sub-safe" environment, ultimately leading to irreversible cumulative damage.

[0004] To compensate for the shortcomings of fixed threshold methods, some art museums have introduced statistical analysis or empirical rule-based methods in their environmental monitoring processes. For example, they use moving averages, daily averages, or periodic statistical analysis of environmental parameters, combined with human experience to make comprehensive judgments about environmental conditions. This can smooth out the impact of short-term fluctuations to some extent. However, their analysis is mostly based on linear statistical characteristics, making it difficult to characterize the non-linear response characteristics of oil painting materials to environmental changes. Furthermore, these methods rely on human experience for risk interpretation, are highly subjective, and the assessment results are difficult to standardize and quantify, and still lack the ability to predict future environmental trends. Some environmental management systems use rule bases or expert systems, pre-setting numerous "if-then" judgment rules to identify risks in different combinations of environmental parameters. While these systems have some applicability in specific scenarios, their rule systems heavily rely on prior knowledge and cannot cover the complex change patterns that emerge during the long-term evolution of environmental conditions. When there are multiple factors coupled or slow drifts among environmental parameters, the rule base often fails to identify potential risks in a timely manner, and the system expansion and maintenance costs increase significantly with the number of rules. Summary of the Invention

[0005] To address the technical problem mentioned above, which focuses solely on the instantaneous state of environmental parameters at a single moment during the risk assessment of oil painting preservation environments, neglecting the evolutionary trends of environmental conditions over time and the cumulative and irreversible nature of oil painting material deterioration, thus making it difficult to identify long-term potential risks in a timely manner, this technical solution provides a method for assessing the environmental risks of oil painting preservation. This method transforms the traditional passive alarm mode into an active early warning mode, significantly improving the accuracy of risk identification, the stability of assessment, and the scientific and reliable nature of conservation decisions during the long-term preservation of oil paintings; it effectively solves the aforementioned technical problems.

[0006] This invention is achieved through the following technical solution:

[0007] A method for risk assessment of the preservation environment of oil paintings, including the following steps:

[0008] Step 1: Collect environmental data and construct a time series;

[0009] Step 2: Preprocess the environmental data collected in Step 1 to obtain environmental time series data; then slice the environmental time series data using a fixed time window sliding method to construct feature samples for input to the time series prediction model;

[0010] Step 3: Construct an improved time series prediction model. The improved time series prediction model is based on a long short-term memory neural network (LSTM) and its memory evolution mechanism is improved to meet the risk assessment needs of the oil painting preservation environment. After the prediction model is constructed, the environmental parameters of the oil painting are predicted based on the improved time series prediction model, and the prediction results are output.

[0011] Step 4: Based on the prediction results output in Step 3, model and analyze the potential long-term impact of the predicted environmental conditions on oil painting materials to achieve early identification and quantitative assessment of environmental risks to oil painting preservation. The assessment focuses on depicting the continuous and cumulative impact of environmental factors on oil paintings over time, and more realistically reflects the deterioration process of oil painting materials aging, fading, cracking, and contamination adsorption.

[0012] Furthermore, the specific operation method for collecting environmental data in step one is as follows: Multiple environmental sensors are deployed within the oil painting exhibition space to monitor the micro-environmental parameters surrounding the oil paintings over a long period. The environmental parameters collected by these sensors are collected according to a unified time reference. The environmental sensors include those for collecting temperature, relative humidity, light intensity, ultraviolet radiation intensity, and air pollutant concentration. A temperature sensor is used to obtain changes in the air temperature around the oil paintings. A relative humidity sensor reflects changes in the moisture content of the air. A light intensity sensor monitors the intensity of visible light irradiating the oil painting surface. An ultraviolet radiation sensor monitors the level of ultraviolet radiation that may cause pigment fading and material aging. An air pollutant sensor detects the concentration of pollutants in the air that have a potential corrosive effect on the oil painting material. Day / night indicator data from the sensors is automatically generated based on the collection time information and is used to distinguish the environmental state under natural light conditions from those under non-light conditions.

[0013] Furthermore, the specific operation method of step one is as follows: the construction of the time series in step one is specifically as follows: the sampling time interval is set to a fixed time interval. Each sampling time corresponds to a set of environmental state vectors It can be represented as:

[0014] (1);

[0015] in, Indicates time Ambient temperature, Indicates relative humidity. Indicates light intensity. Indicates the intensity of ultraviolet radiation. Indicates the concentration of air pollutants. Indicates the day / night indicator variable;

[0016] The continuously collected environmental state vectors are arranged in chronological order to construct a multidimensional time-series dataset of the oil painting preservation environment. :

[0017] (2);

[0018] in, This indicates the total number of time steps for data collection.

[0019] Furthermore, the environmental data preprocessing described in step two is specifically performed as follows:

[0020] Step 2.1: For missing data, adopt the corresponding data imputation strategy according to the missing situation; the data imputation strategy is as follows: when the missing time is short and the data before and after the time point is complete, use the interpolation method based on time continuity to imput the missing data; when the missing time is long, combine the historical data of the same period or the changing trend of environmental parameters to estimate and imput the data.

[0021] Step 2.2: Perform anomaly detection and correction on the collected data; when an anomaly is detected, correct it by using methods such as smoothing replacement of data from nearby time points, moving average correction, or limiting it to a reasonable range;

[0022] Step 2.3: Normalize the environmental parameters to map data with different dimensions to a unified numerical range, thereby improving the convergence speed and prediction accuracy of model training. All environmental parameters should have consistent numerical scales.

[0023] Furthermore, the specific operation method for constructing feature samples in step two is as follows: setting the time window length to... That is, each feature sample contains continuous Environmental parameter data within a time step; by sliding the time window on the time axis with a preset step size, multiple sets of continuous environmental state data are sequentially extracted to form a feature sample set; each feature sample It can be represented as:

[0024] (3);

[0025] in, Indicates time A multidimensional environment parameter vector;

[0026] A single feature sample contains environmental state information at the current moment and integrates environmental change processes from multiple historical moments, enabling subsequent time series prediction models to fully learn the temporal dependencies and trends of environmental parameters.

[0027] Furthermore, in step three, the environmental parameters of the oil painting are predicted. In the time series prediction process, the model uses the environmental parameters within multiple consecutive time steps as input to describe the change process of the oil painting's preservation environment over a period of time.

[0028] Set at time step The environment parameter vector at time t is By combining environmental parameter vectors from multiple consecutive moments in chronological order to form an input sequence, the model can simultaneously perceive the historical state and direction of change of environmental parameters, thus providing a sufficient information basis for subsequent predictions; Long Short-Term Memory Neural Networks introduce memory units in the time dimension. With hidden state This allows for the storage and updating of historical environmental parameter information, enabling the modeling of long-term environmental evolution patterns; at each time step... LSTM networks selectively memorize historical information through input gates, forget gates, and output gates. Their basic update process is represented as follows:

[0029] (4);

[0030] in, Indicates time step The hidden state; Indicates time step Historical unit status; These are the outputs of the forget gate and the input gate, respectively. This represents candidate memory information generated jointly by the current environmental parameter input and the historical hidden state; Indicates time The memory unit state is used to store long-term evolution information of the oil painting's preservation environment over time; , The trainable parameter matrix and bias terms; and These represent the Sigmoid function and the hyperbolic tangent function, respectively. This indicates element-wise multiplication.

[0031] Furthermore, the construction of the improved time series prediction model described in step three is specifically carried out as follows: based on the LSTM network model, an accumulated environmental change is introduced in the time dimension to characterize the persistence of environmental changes over a period of time. The accumulated environmental change... Defined as:

[0032] (5);

[0033] in, for The environment parameter vector at time, for The environmental parameter vector at time; Equation (5) represents the cumulative amount of environmental change used to distinguish between the following two situations: occasional drastic changes in a short period of time; and continuous, slow but consistent environmental drift over a long period of time.

[0034] To characterize the stability of the direction of environmental change, an environmental change consistency factor is introduced. :

[0035] (6);

[0036] Among them, consistency factor This factor is used to measure whether environmental changes have similar directions of change in adjacent time periods; when environmental changes exhibit long-term unidirectional drift, this factor takes a larger value; when the direction of environmental changes frequently reverses, this factor takes a smaller value.

[0037] Taking into account the cumulative magnitude, persistence, and directional stability of environmental changes, a memory decay regulation factor is constructed, expressed as:

[0038] (7);

[0039] in, This is a mapping function used to restrict the control factor within a preset range; These are weight parameters; It serves as a regulatory factor, used to dynamically control the degree of retention of historical memory;

[0040] To make the memory renewal process more consistent with the physical characteristics of the long-term evolution of the oil painting preservation environment, an environment-evolution-driven memory decay regulation factor is introduced on top of the basic structure. The strength of historical memory information retention is further adjusted, and the updated memory units are... and hidden state Represented as:

[0041] (8);

[0042] Through Equation (8), the model can strengthen the memory of historical environmental states when environmental changes are slow and trends are stable; when the environment changes abruptly or fluctuates randomly, it can automatically reduce the impact of historical information on current predictions and improve the stability and physical rationality of prediction results.

[0043] Based on the updated memory units With hidden state The model outputs predictions of environmental parameters for future time steps. Represented as:

[0044] (9);

[0045] in, It is a bias term;

[0046] The prediction results not only reflect the instantaneous changing trends of environmental parameters, but also imply the long-term evolution characteristics of the environment, providing highly reliable predictive input for subsequent assessment of cumulative damage to oil paintings and detection of environmental risks.

[0047] Furthermore, the specific operation method of step four is as follows:

[0048] Step 4.1: Based on the output of Step 3, obtain the sequence of predicted environmental parameters arranged in chronological order within the future prediction time interval:

[0049] (10);

[0050] Each predicted environment parameter vector is defined as follows:

[0051] (11);

[0052] This predicted environmental parameter sequence is used to describe the overall environmental evolution trend of the oil painting preservation space over a period of time, rather than the isolated state at a single moment, providing a temporally continuous input basis for subsequent damage modeling.

[0053] Step 4.2: Quantitatively describe the deviation of the predicted environmental parameters from the ideal preservation conditions. For any environmental parameter... Construct its deviation function , represented as:

[0054] (12);

[0055] in, This indicates the recommended target values ​​for the long-term preservation of oil paintings; This indicates the allowable normal fluctuation range of this environmental parameter; It is the first The deviation function value of each predicted environmental parameter; It is the predicted number A predicted environmental parameter; this deviation function reflects the degree of relative deviation of the environmental parameter from the ideal state, providing a unified measurement basis for parameters of different dimensions;

[0056] Step 4.3: Introduce a nonlinear environmental stress intensity function for any environmental parameter. The intensity of its environmental stress is defined as:

[0057] (13);

[0058] in, It is the first The predicted environmental stress intensity value of each environmental parameter; It is the first The deviation function value of each predicted environmental parameter; The index is used to describe the differences in the sensitivity of oil paintings to different environmental factors, enabling the model to reflect the nonlinear transition characteristics between "slow accumulation" and "rapid deterioration"; within each prediction time step, the impact of environmental conditions on oil painting materials is regarded as an instantaneous damage effect.

[0059] Step 4.4: Based on the multidimensional environmental stress intensity, the instantaneous damage increment is defined as:

[0060] (14);

[0061] in, It is the first The instantaneous damage increment value; weighting coefficient It is used to reflect the relative contribution of different environmental factors to the deterioration process of oil paintings; this instantaneous damage increment only reflects the impact of a single time step and cannot directly represent the overall preservation status of the oil painting.

[0062] Step 4.5: Construct a cumulative damage evolution model based on the time memory effect, defined in time... The cumulative damage index of oil paintings is Their evolutionary relationship is as follows:

[0063] (15);

[0064] in, The cumulative damage index is used to characterize the potential degree of degradation of oil paintings under long-term environmental exposure. This is a damage memory retention factor, used to regulate the persistence intensity of historical damage in the current stage; the cumulative damage from the previous moment. It will not disappear automatically as time goes on, but will continue to affect the current state at a certain ratio;

[0065] Step 4.6: Dynamically correlate the impairment memory retention factor with the predicted magnitude of environmental change:

[0066] (16);

[0067] in, The non-negative adjustment coefficient is used to control the strength of the effect of the predicted magnitude of environmental change on the damage memory retention factor; when When the impact of environmental changes is significant, the inhibitory effect on historical traumatic memories is more pronounced; when... When the age is smaller, the sensitivity of impaired memory to environmental changes is reduced;

[0068] Step 4.7: Based on the changes in the cumulative damage index within the predicted time interval, classify and determine the risk level of the oil painting preservation environment; specifically, the risk level is defined as:

[0069] (17);

[0070] in, These represent the lower and upper limits of the cumulative damage index, respectively. Representing the risk level, it is used to comprehensively reflect the potential threat of environmental conditions to the long-term preservation safety of oil paintings over a period of time. When the predicted environmental risk level reaches medium or high risk, the system generates targeted environmental control suggestions or instructions based on the type of environmental parameters from which the risk originates, including measures such as temperature and humidity regulation, light restriction, ultraviolet shielding, and air purification.

[0071] Beneficial effects

[0072] The oil painting preservation environment risk assessment method proposed in this invention has the following advantages compared with the prior art:

[0073] (1) This technical solution unifies the historical monitoring data of multiple environmental parameters such as temperature, relative humidity, light intensity, ultraviolet radiation, and air pollutants in the oil painting preservation environment to construct a time series of multi-dimensional environmental parameters with temporal continuity. Based on this, the trend of environmental state changes in the future period is predicted. The predicted environmental parameters are then introduced into the damage mechanism analysis process of oil painting materials to construct a damage evolution model that can reflect the long-term cumulative effect of environmental action. This transforms the instantaneous fluctuations of environmental parameters into a quantitative impact assessment on the safety of oil painting preservation, realizing the forward-looking identification and classification of environmental risks in oil painting preservation. This changes the risk assessment of oil painting preservation environment from the traditional passive alarm mode to an active early warning mode, significantly improving the accuracy of risk identification, the stability of assessment, and the scientificity and reliability of protection decisions in the long-term preservation of oil paintings.

[0074] (2) This technical solution transforms environmental management from passive response to early warning by trend prediction and comprehensive evaluation of environmental parameters, enabling corresponding measures to be taken before environmental anomalies have a substantial impact on oil paintings. By analyzing and predicting the time series of environmental parameters such as temperature, humidity, and light, it is possible to determine in advance whether the environment is trending in an unfavorable direction, thereby reducing the risk of oil paintings being exposed to adverse environments for a long time and improving the overall stability of the preservation environment.

[0075] (3) This technical solution further introduces the concept of cumulative environmental impact assessment, comprehensively calculating the changes in environmental parameters over a period of time, rather than making judgments based on monitoring data at a single moment. This approach effectively avoids the problem of frequent alarm triggering due to short-term fluctuations, making the risk assessment results more in line with the actual needs of long-term preservation of oil painting materials. In the environmental risk assessment process, this invention fully considers the continuous impact of historical environmental conditions on the state of oil paintings, enabling the environmental assessment results to reflect the potential damage that oil paintings may suffer under long-term environmental influences. Through this approach, the safety level of the oil painting preservation environment can be described more realistically, providing managers with more valuable reference points for judgment. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0077] Figure 2 This is a schematic diagram showing the distribution of prediction errors between the predicted environmental parameters obtained by the method in this invention and the actual monitored values.

[0078] Figure 3 This is a graph showing the change in the loss function of the time series prediction model during the training process in this invention.

[0079] Figure 4 This invention presents the prediction results of temperature and relative humidity parameters in the oil painting preservation environment, as well as the corresponding environmental risk level distribution. Detailed Implementation

[0080] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. The described embodiments are merely some embodiments of the present invention, and not all embodiments. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the design concept of the present invention should fall within the protection scope of the present invention.

[0081] Example 1:

[0082] A method for assessing the environmental risks of oil painting preservation, specifically involving a method based on time series prediction and cumulative damage modeling; including the following steps:

[0083] Step 1: Collect environmental data and construct a time series;

[0084] To comprehensively and continuously obtain information about the environmental conditions for oil painting preservation, various environmental sensors are deployed in the art museum's oil painting exhibition space to monitor the micro-environmental parameters surrounding the paintings over a long period. The environmental parameters collected by these sensors are based on a unified time frame. These sensors include those for collecting temperature, relative humidity, light intensity, ultraviolet radiation intensity, and air pollutant concentrations. Temperature sensors are used to obtain data on changes in the air temperature around the paintings; relative humidity sensors reflect changes in the moisture content of the air; light intensity sensors monitor the intensity of visible light irradiating the oil painting surface; ultraviolet radiation sensors monitor the level of ultraviolet radiation that may cause pigment fading and material aging; and air pollutant sensors detect the concentration of pollutants in the air that have a potential corrosive effect on the oil painting materials. The sensors also have day / night indicator data, automatically generated based on the collection time information, used to distinguish between environmental conditions under natural light and non-light conditions.

[0085] All environmental parameters are collected according to a unified time reference. In this embodiment, the sampling time interval is set to a fixed time interval. For example, data is collected once per hour, thus forming a continuous, equally spaced environmental monitoring data sequence; each sampling moment corresponds to a set of environmental state vectors. It can be represented as:

[0086] (1);

[0087] in, Indicates time Ambient temperature, Indicates relative humidity. Indicates light intensity. Indicates the intensity of ultraviolet radiation. Indicates the concentration of air pollutants. This represents the day / night indicator variable.

[0088] The continuously collected environmental state vectors are arranged in chronological order to construct a multidimensional time-series dataset of the oil painting preservation environment. :

[0089] (2);

[0090] in, This indicates the total number of time steps for data collection.

[0091] The multidimensional time series data constructed in the above manner can not only reflect the state characteristics of the oil painting preservation environment at a single moment, but also describe the trend, periodicity and abrupt change characteristics of environmental parameters over time, providing basic data support for subsequent environmental evolution prediction and cumulative damage modeling.

[0092] Step Two: Preprocess the environmental data collected in Step One to obtain environmental time series data; then, slice the environmental time series data using a sliding method with a fixed time window to construct feature samples for input to the time series prediction model; the specific operation method is as follows:

[0093] Step 2.1: For missing data, adopt the corresponding data imputation strategy according to the missing situation; the data imputation strategy is as follows: when the missing time is short and the data before and after the time point is complete, use the interpolation method based on time continuity to imput the missing data; when the missing time is long, combine the historical data of the same period or the changing trend of environmental parameters to estimate and imput the data.

[0094] Step 2.2: Perform anomaly detection and correction on the collected data; when an anomaly is detected, correct it by using methods such as smoothing replacement of data from nearby time points, moving average correction, or limiting it to a reasonable range;

[0095] Step 2.3: Normalize the environmental parameters to map data with different dimensions to a unified numerical range, thereby improving the convergence speed and prediction accuracy of model training. All environmental parameters should have consistent numerical scales.

[0096] Step 2.4: After completing the data preprocessing operations, construct feature samples; specifically, set the time window length to... That is, each feature sample contains continuous Environmental parameter data within a time step; by sliding the time window on the time axis with a preset step size, multiple sets of continuous environmental state data are sequentially extracted to form a feature sample set; each feature sample It can be represented as:

[0097] (3);

[0098] in, Indicates time A multidimensional environment parameter vector;

[0099] Through the above feature construction method, a single feature sample not only contains the environmental state information at the current moment, but also integrates the environmental change process at multiple historical moments, so that the subsequent time series prediction model can fully learn the time dependence and change trend of environmental parameters.

[0100] Step 3: Construct an improved time series prediction model. The improved time series prediction model is based on a long short-term memory neural network (LSTM) and its memory evolution mechanism is improved to meet the risk assessment needs of the oil painting preservation environment. After the prediction model is constructed, the environmental parameters of the oil painting are predicted based on the improved time series prediction model, and the prediction results are output.

[0101] After completing the environmental data preprocessing and feature construction, the constructed time series feature samples are input into the improved time series prediction model to predict the changing trends of the oil painting preservation environment parameters in the future time range.

[0102] In time series forecasting, the model uses environmental parameters from multiple consecutive time steps as input to describe the changes in the oil painting preservation environment over a period of time; let's assume that at time step... The environment parameter vector at time t is By combining environmental parameter vectors from multiple consecutive moments in chronological order to form an input sequence, the model can simultaneously perceive the historical state and direction of change of environmental parameters, thus providing a sufficient information foundation for subsequent predictions. Long Short-Term Memory (LSTM) neural networks introduce memory units in the time dimension. With hidden state This allows for the storage and updating of historical environmental parameter information, enabling the modeling of long-term environmental evolution patterns; at each time step... LSTM networks selectively memorize historical information through input gates, forget gates, and output gates. Their basic update process is represented as follows:

[0103] (4);

[0104] in, Indicates time step The hidden state; Indicates time step Historical unit status; These are the outputs of the forget gate and the input gate, respectively. This represents candidate memory information generated jointly by the current environmental parameter input and the historical hidden state; Indicates time The memory unit state is used to store long-term evolution information of the oil painting's preservation environment over time; , The trainable parameter matrix and bias terms; and These represent the Sigmoid function and the hyperbolic tangent function, respectively. This indicates element-wise multiplication.

[0105] Through the memory update mechanism described in formula (4), the LSTM network can store and update the historical information of the environmental parameters of the oil painting in the time dimension, thereby learning the long-term dependence of environmental parameter changes and providing a basic model structure for subsequent environmental state prediction.

[0106] The effects of the preservation environment on the material degradation of oil paintings are characterized by slow accumulation, irreversibility, and a response different from short-term, drastic fluctuations. In traditional time series models, the retention or forgetting of historical memory usually depends only on the current input state, failing to reflect the continuity and directionality of environmental changes over time.

[0107] This paper proposes a memory decay regulation mechanism based on environmental evolution inertia. By simultaneously considering the magnitude, duration, and directional stability of environmental changes, it jointly regulates the decay rate of historical memory through multiple factors, thereby making the model's memory evolution mode more closely reflect the actual physical characteristics of the oil painting preservation environment. An environmental change accumulation rate is introduced in the time dimension to characterize the persistence of environmental changes over a period of time. Defined as:

[0108] (5);

[0109] in, for The environment parameter vector at time, for The environmental parameter vector at time; Equation (5) represents the cumulative amount of environmental change used to distinguish between the following two situations: occasional drastic changes in a short period of time; and continuous, slow but consistent environmental drift over a long period of time.

[0110] Furthermore, to characterize the stability of the direction of environmental change, an environmental change consistency factor is introduced. :

[0111] (6);

[0112] Among them, consistency factor This factor is used to measure whether environmental changes have similar directions of change in adjacent time periods; when environmental changes exhibit long-term unidirectional drift, this factor takes a larger value; when the direction of environmental changes frequently reverses, this factor takes a smaller value.

[0113] Taking into account the cumulative magnitude, persistence, and directional stability of environmental changes, a memory decay regulation factor is constructed, expressed as:

[0114] (7);

[0115] in, This is a mapping function used to restrict the control factor within a preset range; These are weight parameters; It serves as a regulatory factor, used to dynamically control the degree of retention of historical memory.

[0116] In the LSTM basic memory update mechanism described in formula (4), the historical cell state The degree of decay is determined solely by the forgetting gate. The forgetting gate's output, while controlled by the current time step, primarily depends on the environmental input state and fails to reflect the persistence and cumulative nature of environmental parameter changes over time. To make the memory update process more consistent with the physical characteristics of the long-term evolution of the oil painting preservation environment, an environmental evolution-driven memory decay regulation factor is introduced on top of the basic structure. The strength of historical memory information retention is further adjusted, and the updated memory units are... and hidden state Represented as:

[0117] (8);

[0118] Through Equation (8), the model can strengthen the memory of historical environmental states when environmental changes are slow and trends are stable; when the environment changes abruptly or fluctuates randomly, it can automatically reduce the impact of historical information on current predictions and improve the stability and physical rationality of prediction results.

[0119] Based on the updated memory units With hidden state The model outputs predictions of environmental parameters for future time steps. Represented as:

[0120] (9);

[0121] in, It is a bias term;

[0122] The prediction results not only reflect the instantaneous changing trends of environmental parameters, but also imply the long-term evolution characteristics of the environment, providing highly reliable predictive input for subsequent assessment of cumulative damage to oil paintings and detection of environmental risks.

[0123] Step Four: Based on the prediction results output in Step Three, model and analyze the potential long-term impact of the predicted environmental conditions on oil painting materials to achieve early identification and quantitative assessment of environmental risks to oil painting preservation. The assessment focuses on depicting the continuous and cumulative effects of environmental factors on oil paintings over time, more realistically reflecting the deterioration process of oil painting materials, including aging, fading, cracking, and contamination adsorption. The specific operation method is as follows:

[0124] Step 4.1: Based on the output of Step 3, obtain the sequence of predicted environmental parameters arranged in chronological order within the future prediction time interval:

[0125] (10);

[0126] Each predicted environment parameter vector is defined as follows:

[0127] (11);

[0128] This predicted environmental parameter sequence is used to describe the overall environmental evolution trend of the oil painting preservation space over a period of time, rather than the isolated state at a single moment, providing a temporally continuous input basis for subsequent damage modeling.

[0129] Step 4.2: Considering that there is no single, absolutely "safe" or "dangerous" environmental threshold for oil painting preservation, but rather that fluctuations are allowed within a certain range, we first quantitatively describe the degree of deviation of environmental parameters from ideal preservation conditions. For any given environmental parameter... Construct its deviation function , represented as:

[0130] (12);

[0131] in, This indicates the recommended target values ​​for the long-term preservation of oil paintings; This indicates the allowable normal fluctuation range of this environmental parameter; It is the first The deviation function value of each predicted environmental parameter; It is the predicted number A predicted environmental parameter; this deviation function reflects the degree of relative deviation of the environmental parameter from the ideal state, providing a unified measurement basis for parameters of different dimensions.

[0132] Step 4.3: Because oil painting materials exhibit a significantly nonlinear response to environmental changes, slight deviations may have almost no impact, while sustained or large deviations will significantly accelerate the degradation process. Therefore, a nonlinear environmental stress intensity function is further introduced. For any environmental parameter... The intensity of its environmental stress is defined as:

[0133] (13);

[0134] in, It is the first The predicted environmental stress intensity value of each environmental parameter; It is the first The deviation function value of each predicted environmental parameter; The index is used to describe the varying sensitivities of oil paintings to different environmental factors, enabling the model to reflect the non-linear transition between "slow accumulation" and "rapid deterioration." Within each prediction time step, the impact of environmental conditions on the oil painting material is considered an instantaneous damage effect. This effect is not directly equivalent to damage to the oil painting, but rather represents the accumulation of a potential adverse impact.

[0135] Step 4.4: Based on the multidimensional environmental stress intensity, the instantaneous damage increment is defined as:

[0136] (14);

[0137] in, It is the first The instantaneous damage increment value; weighting coefficient This is used to reflect the relative contribution of different environmental factors to the degradation process of oil paintings; for example, the influence of ultraviolet radiation on pigment fading has a higher weight than the short-term effects of temperature fluctuations. This instantaneous damage increment only reflects the effect of a single time step and cannot directly represent the overall preservation status of the oil painting.

[0138] Step 4.5: Since the degradation process of oil painting materials exhibits a significant time memory characteristic, meaning that historical environmental conditions continue to influence the current state, a cumulative damage evolution model based on the time memory effect is constructed, defined over time... The cumulative damage index of oil paintings is Their evolutionary relationship is as follows:

[0139] (15);

[0140] in, The cumulative damage index is used to characterize the potential degree of degradation of oil paintings under long-term environmental exposure. This is a damage memory retention factor, used to regulate the persistence intensity of historical damage in the current stage; the cumulative damage from the previous moment. It will not disappear automatically as time goes on, but will continue to affect the current state at a certain ratio.

[0141] Step 4.6: Considering the impact of environmental changes on the rate of damage accumulation, dynamically correlate the damage memory retention factor with the predicted magnitude of environmental changes:

[0142] (16);

[0143] in, The non-negative adjustment coefficient is used to control the strength of the effect of the predicted magnitude of environmental change on the damage memory retention factor; when When the impact of environmental changes is significant, the inhibitory effect on historical traumatic memories is more pronounced; when... When the age is small, the sensitivity of damaged memories to environmental changes is reduced.

[0144] Step 4.7: Based on the changes in the cumulative damage index within the predicted time interval, classify and determine the risk level of the oil painting preservation environment; specifically, the risk level is defined as:

[0145] (17);

[0146] in, These represent the lower and upper limits of the cumulative damage index, respectively. Representing the risk level, it is used to comprehensively reflect the potential threat of environmental conditions to the long-term preservation safety of oil paintings over a period of time. When the predicted environmental risk level reaches medium or high risk, the system generates targeted environmental control suggestions or instructions based on the type of environmental parameters from which the risk originates, including measures such as temperature and humidity regulation, light restriction, ultraviolet shielding, and air purification.

[0147] To verify the effectiveness of the proposed method for detecting environmental risks in oil painting preservation based on time series prediction and cumulative environmental effects, an experimental environment was constructed and verification experiments were conducted. The experiments were carried out in a general-purpose computer environment. The experimental platform possesses conventional data processing and model training capabilities, meeting the training and inference requirements of the time series prediction model.

[0148] The software environment used in the experiment supports the entire process of data preprocessing, model building, training, and prediction, and can stably run long-term series data analysis tasks. During the experiment, it did not rely on dedicated equipment or systems from specific vendors, demonstrating that the method of this invention has good versatility and deployability, and is applicable to different art museums or cultural relic preservation scenarios.

[0149] The experimental data comes from actual environmental monitoring data of the exhibition or storage environment of oil paintings in art museums. The data covers a complete natural year cycle and can fully reflect the variation characteristics of environmental parameters under different seasons and diurnal conditions. The dataset used has the following characteristics:

[0150] (1) The data covers the entire year of 2023 and is sampled hourly, which can reflect the long-term trend and periodic characteristics of environmental parameters.

[0151] (2) The data includes temperature data, relative humidity data, light intensity data, ultraviolet radiation intensity data, air pollutant concentration data, and day and night identification information.

[0152] (3) The data includes cumulative damage indicators, risk event markers and environmental control behavior records, providing a basis for verifying the risk assessment and early warning mechanism of the present invention.

[0153] The experimental process was carried out sequentially according to the technical flow of the method of this invention, mainly including the following steps:

[0154] (1) Preprocess the original environmental time series data, including missing value completion, outlier correction and data normalization, to eliminate the influence of sensor noise and occasional interference on the experimental results.

[0155] (2) By using a sliding time window approach, environmental parameters from multiple consecutive time steps are combined into feature samples to construct the input dataset for the time series prediction model. This process can preserve the correlation between environmental parameters in the time dimension.

[0156] (3) Based on the constructed feature samples, train an environmental parameter prediction model to predict the future trend of environmental parameter changes. The prediction results serve as an important input for subsequent environmental risk assessment. Combine the predicted environmental parameters with historical environmental effects to calculate the cumulative impact of environmental changes over time, thereby forming a comprehensive environmental risk assessment result.

[0157] (4) Determine the risk status of the oil painting preservation environment based on the assessment results and output the corresponding risk level to verify the application effect of the method of the present invention in actual environmental management.

[0158] Figure 2 This diagram illustrates the distribution of prediction errors between the environmental parameter predictions obtained using the method of this invention and the actual monitored values. It reflects the overall prediction accuracy and error stability of the time series prediction model at different time steps. As can be seen from the diagram, most prediction errors are concentrated near zero, with small error amplitudes, and the overall distribution exhibits a relatively symmetrical and concentrated characteristic, indicating a high degree of consistency between the model's predictions and the actual environmental change trends.

[0159] Further analysis revealed that the prediction error did not exhibit a significant long-tail distribution or extreme bias points, indicating that the improved time series prediction model employed in this invention maintains good robustness in the face of periodic changes and slow drifts in environmental parameters, and is insensitive to short-term noise and occasional fluctuations. This result demonstrates that by introducing a memory regulation mechanism driven by environmental evolution characteristics, the model avoids the continuous accumulation of errors during long-term prediction, providing a reliable data foundation for subsequent cumulative damage assessment based on the prediction results.

[0160] Figure 3 The graph shows the loss function variation curve of the time series prediction model during training, reflecting the convergence process and stability of the model training. As can be observed from the graph, with the increase of training rounds, the model loss value generally shows a gradual decrease and eventually stabilizes, without obvious oscillations or divergence, indicating that the model training process is stable and the parameter update direction is reasonable. In the early stage of training, the loss value decreases rapidly, indicating that the model can quickly learn the basic changing patterns of the oil painting preservation environment parameters over time. In the later stage of training, the loss value gradually converges and remains at a low level, indicating that the model has sufficiently captured the time dependence and long-term evolution characteristics between environmental parameters.

[0161] The results verify the rationality of introducing an environmental evolution regulation mechanism on the basis of the standard LSTM structure in this invention. This improves the model's ability to predict long-term trends without increasing training instability, and provides a stable and reliable predictive model support for the forward-looking assessment of environmental risks.

[0162] Figure 4 This figure shows the prediction results of temperature and relative humidity parameters in the oil painting preservation environment based on the method of this invention, as well as the corresponding distribution of environmental risk levels. The figure also illustrates the predicted change curves of environmental parameters and the evolution of risk levels over time, visually reflecting the correspondence between environmental change trends and risk assessment results. The temperature and humidity prediction curves show that the environmental parameters generally exhibit a slow changing trend over time, superimposed with a certain degree of periodic fluctuation. In some time intervals, although the predicted temperature or humidity values ​​do not significantly exceed the traditional safety threshold range, their trend of continuously deviating from ideal preservation conditions is quite obvious. Based on the cumulative damage modeling method proposed in this invention, the system outputs the corresponding environmental risk assessment results at this stage.

[0163] By comparison, it can be seen that the method of this invention does not only trigger risk warnings when environmental parameters instantaneously exceed the standard, but can comprehensively consider the predicted trend of environmental changes and their cumulative impact over time, and identify long-term potential risks in advance. The cumulative damage index and risk classification mechanism constructed based on the prediction results can effectively avoid the misjudgment problem of "safety if not exceeding the standard" in traditional threshold methods, making the risk assessment results more in line with the actual needs of long-term preservation of oil painting materials.

[0164] comprehensive Figures 2 to 4 The experimental results show that the oil painting preservation environment risk assessment method based on time series prediction and cumulative damage modeling proposed in this invention can achieve a comprehensive analysis of environmental change trends and their long-term impacts while ensuring prediction accuracy and model stability, providing effective technical support for the forward-looking risk management of oil painting preservation environments.

[0165] Analysis of the experimental results shows that the predicted environmental parameter change curves are highly consistent with the actual environmental change trends, reflecting the overall trend of environmental parameters during diurnal and seasonal changes, rather than just responding to short-term fluctuations. Experimental results indicate that relying solely on a single-moment environmental threshold can easily overlook the risks arising from long-term deviations from the ideal environmental range; however, this invention, through a cumulative assessment mechanism, can identify environmental changes that are continuously in a sub-risk state. In the experimental samples, when environmental parameters have not yet significantly exceeded the safety threshold, the method of this invention has already identified potential risks through predicted trends and cumulative effects, thus providing early risk warnings. Even when environmental parameters fluctuate frequently in the short term, the risk results output by the method of this invention do not show frequent jumps, indicating that the method has a certain robustness to noise and occasional changes. Experimental results show that the oil painting preservation environment risk detection method based on time series prediction and environmental cumulative effects proposed in this invention can effectively integrate multi-dimensional environmental parameter information, accurately depicting the changing trends of the oil painting preservation environment and its long-term impact.

[0166] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention, or modify them into equivalent embodiments, without departing from the scope of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the present invention, should fall within the protection scope of the present invention.

Claims

1. A method for risk assessment of the preservation environment of oil paintings, characterized in that: Including the following steps: Step 1: Collect environmental data and construct a time series; Step 2: Preprocess the environmental data collected in Step 1 to obtain environmental time series data; then slice the environmental time series data using a fixed time window sliding method to construct feature samples for input to the time series prediction model; Step 3: Construct an improved time series prediction model. The improved time series prediction model is based on a long short-term memory neural network (LSTM) and its memory evolution mechanism is improved to meet the risk assessment needs of the oil painting preservation environment. After the prediction model is constructed, the environmental parameters of the oil painting are predicted based on the improved time series prediction model, and the prediction results are output. Step 4: Based on the prediction results output in Step 3, model and analyze the potential long-term impact of the predicted environmental conditions on oil painting materials to achieve early identification and quantitative assessment of environmental risks to oil painting preservation. The assessment focuses on depicting the continuous and cumulative impact of environmental factors on oil paintings over time, and more realistically reflects the deterioration process of oil painting materials aging, fading, cracking, and contamination adsorption.

2. The method for risk assessment of oil painting preservation environment according to claim 1, characterized in that: The specific operation method for collecting environmental data in step one is as follows: Multiple environmental sensors are deployed within the oil painting exhibition space to monitor the micro-environmental parameters surrounding the oil paintings over a long period. The environmental parameters collected by these sensors are collected according to a unified time reference. The environmental sensors include those for collecting temperature, relative humidity, light intensity, ultraviolet radiation intensity, and air pollutant concentration. A temperature sensor is used to obtain changes in the air temperature around the oil paintings. A relative humidity sensor reflects changes in the moisture content of the air. A light intensity sensor monitors the intensity of visible light irradiating the oil painting surface. An ultraviolet radiation sensor monitors the level of ultraviolet radiation that may cause pigment fading and material aging. An air pollutant sensor detects the concentration of pollutants in the air that have a potential corrosive effect on the oil painting materials. Day / night indicator data from the sensors is automatically generated based on the collection time information and is used to distinguish the environmental state under natural light conditions from that under non-light conditions.

3. The method for risk assessment of oil painting preservation environment according to claim 1, characterized in that: The specific operation method of step one is as follows: the construction of the time series in step one is specifically as follows: the sampling time interval is set to a fixed time interval. Each sampling time corresponds to a set of environmental state vectors It can be represented as: (1)； in, Indicates time Ambient temperature, Indicates relative humidity. Indicates light intensity. Indicates the intensity of ultraviolet radiation. Indicates the concentration of air pollutants. Indicates the day / night indicator variable; The continuously collected environmental state vectors are arranged in chronological order to construct a multidimensional time-series dataset of the oil painting preservation environment. : (2)； in, This indicates the total number of time steps for data collection.

4. The method for risk assessment of oil painting preservation environment according to claim 1, characterized in that: The specific operation method for preprocessing environmental data as described in step two is as follows: Step 2.1: For missing data, adopt the corresponding data imputation strategy according to the missing situation; the data imputation strategy is as follows: when the missing time is short and the data before and after the time point is complete, use the interpolation method based on time continuity to imput the missing data; when the missing time is long, combine the historical data of the same period or the changing trend of environmental parameters to estimate and imput the data. Step 2.2: Perform anomaly detection and correction on the collected data; when an anomaly is detected, correct it by using methods such as smoothing replacement of data from nearby time points, moving average correction, or limiting it to a reasonable range; Step 2.3: Normalize the environmental parameters to map data with different dimensions to a unified numerical range, thereby improving the convergence speed and prediction accuracy of model training. All environmental parameters should have consistent numerical scales.

5. The method for risk assessment of oil painting preservation environment according to claim 1, characterized in that: The specific operation method for constructing feature samples in step two is as follows: set the time window length to... That is, each feature sample contains continuous Environmental parameter data within a time step; by sliding the time window on the time axis with a preset step size, multiple sets of continuous environmental state data are captured sequentially to form a feature sample set; Each feature sample It can be represented as: (3)； in, Indicates time A multidimensional environment parameter vector; A single feature sample contains environmental state information at the current moment and integrates environmental change processes from multiple historical moments, enabling subsequent time series prediction models to fully learn the temporal dependencies and trends of environmental parameters.

6. The method for risk assessment of oil painting preservation environment according to claim 1, characterized in that: In step three, the environmental parameters of the oil painting are predicted. In the time series prediction process, the model takes the environmental parameters in multiple consecutive time steps as input to describe the change process of the oil painting's preservation environment over a period of time. Set at time step The environment parameter vector at time t is By combining environmental parameter vectors from multiple consecutive moments in chronological order to form an input sequence, the model can simultaneously perceive the historical state and direction of change of environmental parameters, thus providing a sufficient information basis for subsequent predictions; Long Short-Term Memory Neural Networks introduce memory units in the time dimension. With hidden state This allows for the storage and updating of historical environmental parameter information, enabling the modeling of long-term environmental evolution patterns; at each time step... LSTM networks selectively memorize historical information through input gates, forget gates, and output gates. Their basic update process is represented as follows: (4)； in, Indicates time step The hidden state; Indicates time step Historical unit status; These are the outputs of the forget gate and the input gate, respectively. This represents candidate memory information generated jointly by the current environmental parameter input and the historical hidden state; Indicates time The memory unit state is used to store long-term evolution information of the oil painting's preservation environment over time; , The trainable parameter matrix and bias terms; and These represent the Sigmoid function and the hyperbolic tangent function, respectively. This indicates element-wise multiplication.

7. The method for risk assessment of oil painting preservation environment according to claim 6, characterized in that: The construction of the improved time series prediction model described in step three is specifically as follows: Based on the LSTM network model, an accumulated environmental change is introduced into the time dimension to characterize the persistence of environmental changes over a period of time. The accumulated environmental change... Defined as: (5)； in, for The environmental parameter vector at time, for The environmental parameter vector at time; Equation (5) represents the cumulative amount of environmental change used to distinguish between the following two situations: occasional drastic changes in a short period of time; and continuous, slow but consistent environmental drift over a long period of time. To characterize the stability of the direction of environmental change, an environmental change consistency factor is introduced. : (6)； Among them, consistency factor This factor is used to measure whether environmental changes have similar directions of change in adjacent time periods; when environmental changes exhibit long-term unidirectional drift, this factor takes a larger value; when the direction of environmental changes frequently reverses, this factor takes a smaller value. Taking into account the cumulative magnitude, persistence, and directional stability of environmental changes, a memory decay regulation factor is constructed, expressed as: (7)； in, This is a mapping function used to restrict the control factor within a preset range; These are weight parameters; It serves as a regulatory factor, used to dynamically control the degree of retention of historical memory; To make the memory renewal process more consistent with the physical characteristics of the long-term evolution of the oil painting preservation environment, an environment-evolution-driven memory decay regulation factor is introduced on top of the basic structure. The strength of historical memory information retention is further adjusted, and the updated memory units are... and hidden state Represented as: (8)； Through Equation (8), the model can strengthen the memory of historical environmental states when environmental changes are slow and trends are stable; when the environment changes abruptly or fluctuates randomly, it can automatically reduce the impact of historical information on current predictions and improve the stability and physical rationality of prediction results. Based on the updated memory units With hidden state The model outputs predictions of environmental parameters for future time steps. Represented as: (9)； in, It is a bias term; The prediction results not only reflect the instantaneous changing trends of environmental parameters, but also imply the long-term evolution characteristics of the environment, providing highly reliable predictive input for subsequent assessment of cumulative damage to oil paintings and detection of environmental risks.

8. The method for risk assessment of the preservation environment of oil paintings according to claim 1, characterized in that: The specific operation method for step four is as follows: Step 4.1: Based on the output of Step 3, obtain the sequence of predicted environmental parameters arranged in chronological order within the future prediction time interval: (10)； Each predicted environment parameter vector is defined as follows: (11)； This predicted environmental parameter sequence is used to describe the overall environmental evolution trend of the oil painting preservation space over a period of time, rather than the isolated state at a single moment, providing a temporally continuous input basis for subsequent damage modeling. Step 4.2: Quantitatively describe the deviation of the predicted environmental parameters from the ideal preservation conditions. For any environmental parameter... Construct its deviation function , is represented as: (12)； in, This indicates the recommended target values ​​for the long-term preservation of oil paintings; This indicates the allowable normal fluctuation range of this environmental parameter; It is the first The deviation function value of each predicted environmental parameter; It is the predicted number A predicted environmental parameter; this deviation function reflects the degree of relative deviation of the environmental parameter from the ideal state, providing a unified measurement basis for parameters of different dimensions; Step 4.3: Introduce a nonlinear environmental stress intensity function for any environmental parameter. The intensity of its environmental stress is defined as: (13)； in, It is the first The predicted environmental stress intensity value of each environmental parameter; It is the first The deviation function value of each predicted environmental parameter; The index is used to describe the differences in the sensitivity of oil paintings to different environmental factors, enabling the model to reflect the nonlinear transition characteristics between "slow accumulation" and "rapid deterioration"; within each prediction time step, the impact of environmental conditions on oil painting materials is regarded as an instantaneous damage effect. Step 4.4: Based on the multidimensional environmental stress intensity, the instantaneous damage increment is defined as: (14)； in, It is the first The instantaneous damage increment value; weighting coefficient It is used to reflect the relative contribution of different environmental factors to the deterioration process of oil paintings; this instantaneous damage increment only reflects the impact of a single time step and cannot directly represent the overall preservation status of the oil painting. Step 4.5: Construct a cumulative damage evolution model based on the time memory effect, defined in time... The cumulative damage index of oil paintings is Their evolutionary relationship is as follows: (15)； in, The cumulative damage index is used to characterize the potential degree of degradation of oil paintings under long-term environmental exposure. This is a damage memory retention factor, used to regulate the persistence intensity of historical damage in the current stage; the cumulative damage from the previous moment. It will not disappear automatically as time goes on, but will continue to affect the current state at a certain ratio; Step 4.6: Dynamically correlate the impairment memory retention factor with the predicted magnitude of environmental change: (16)； in, The non-negative adjustment coefficient is used to control the strength of the effect of the predicted magnitude of environmental change on the damage memory retention factor; when When the impact of environmental changes is significant, the inhibitory effect on historical traumatic memories is more pronounced; when... When the age is smaller, the sensitivity of impaired memory to environmental changes is reduced; Step 4.7: Based on the changes in the cumulative damage index within the predicted time interval, classify and determine the risk level of the oil painting preservation environment; specifically, the risk level is defined as: (17)； in, These represent the lower and upper limits of the cumulative damage index, respectively. Representing the risk level, it is used to comprehensively reflect the potential threat of environmental conditions to the long-term preservation safety of oil paintings over a period of time. When the predicted environmental risk level reaches medium or high risk, the system generates targeted environmental control suggestions or instructions based on the type of environmental parameters from which the risk originates, including measures such as temperature and humidity regulation, light restriction, ultraviolet shielding, and air purification.

Images