Intelligent monitoring alarm system for intermediate frequency furnace

Abstract

The invention relates to the technical field of monitoring and alarming, and particularly discloses an intelligent monitoring and alarming system for an intermediate frequency furnace, which comprises a setting module for selecting a target element and setting the content range of the target element; the calibration module is used for grouping the set content nodes, selecting the target metal in the group i, and determining the target concentration according to the smoke concentration change condition in the melting process of the target metal; generating coordinate points corresponding to the group i in an (n + 1)-dimensional coordinate system; generating coordinate points corresponding to all the groups, and fitting all the coordinate points to obtain a fitted curve; and the alarm module is used for acquiring the contrast content of the currently smelted metal, acquiring the predicted concentration according to the contrast content, and performing early warning according to the predicted concentration. According to the invention, flue dust standard exceeding early warning of the intermediate frequency furnace is realized, and the body health of operators is protected.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring alarms, and in particular to an intelligent monitoring alarm system for a medium frequency furnace. Background Art

[0002] Medium frequency furnace, also known as medium frequency induction furnace, is widely used for smelting non-ferrous metals and ferrous metals. Compared with other casting equipment, medium frequency induction furnace has the advantages of high thermal efficiency, short smelting time, less alloy element burnout, wide smelting material range, low environmental pollution, and the ability to accurately control the temperature and composition of the molten metal.

[0003] During the metal smelting process, medium-frequency induction heating generates a large amount of harmful gases and fine particles, such as carbon monoxide, sulfur dioxide, nitrogen oxides, and metal oxide dust. These pollutants are more active at high temperatures and easily diffuse with furnace gases into the work area, causing damage to the respiratory, circulatory, and nervous systems of operators. Long-term exposure can lead to serious health problems such as chronic obstructive pulmonary disease and cardiovascular disease. However, existing intelligent monitoring systems primarily focus on equipment safety, such as cooling water temperature, coil thermal state, and furnace vibration, and do not adequately monitor air quality. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent monitoring and alarm system for a medium frequency furnace to solve the following technical problems: The existing intelligent monitoring system mainly focuses on equipment safety such as cooling water temperature, coil thermal state and furnace vibration, and lacks monitoring of air quality.

[0005] The purpose of the present invention can be achieved through the following technical solutions: An intelligent monitoring and alarm system for a medium frequency furnace, comprising: Setting module: select the target element and set the content range of the target element; Calibration module: Set several content nodes within the content range, group the content nodes, select the target metal corresponding to group i, add the target metal into the intermediate frequency furnace for melting, and obtain the smoke concentration emitted by the intermediate frequency furnace in real time. The maximum smoke concentration within the preset monitoring time is used as the target concentration; Generate coordinate points (C1, C2, ..., C n , D), C n represents the content of the target element corresponding to the nth dimension, D represents the target concentration, and n represents the total number of the target elements; Generate coordinate points corresponding to all groups, and perform fitting on all the coordinate points to obtain a fitting curve; Alarm module: obtains the content of the target element in the metal currently being smelted, records it as the comparative content, substitutes the comparative content into the fitting curve to obtain the predicted concentration, and sends an alarm message when the predicted concentration is greater than a preset concentration threshold.

[0006] As a further solution of the present invention: selecting the target element includes: Elements contained in the metal are marked as initial elements, the degree of influence of the initial elements on smoke concentration is determined based on a random forest model and / or a gradient boosting tree, and the initial elements whose influence is greater than a preset influence threshold are used as target elements.

[0007] As a further solution of the present invention: selecting target metals includes: Starting from the starting point of the content range, a number of content nodes are set at preset content intervals, and different target elements correspond to different content intervals; Grouping the content nodes of different target elements, wherein different groups contain at least one different content node, and there is only one content node for one target element in the same group; A target metal is selected, wherein the content of the target element in the target metal is the content corresponding to the content node in group i.

[0008] As a further solution of the present invention: obtaining the smoke concentration includes: A laser scattering monitoring device is installed at a preset position to obtain the smoke concentration based on the reflection of the laser.

[0009] As a further solution of the present invention: the process of obtaining the target concentration further includes: Draw a curve A showing the change of smoke concentration over time during the monitoring period; Draw a straight line through (0, max(f(t))) and parallel to the x-axis, and draw a straight line through (0, max(f(t))-Δd) and parallel to the x-axis, where Δd is a first preset value; The portion of the curve A between the two straight lines is recorded as the normal portion, and the proportion of the domain of the normal portion to the monitoring time is calculated and recorded as the normal proportion; If the normal ratio is less than 0.3, both straight lines are shifted downward by Δd', where Δd' is a second preset value and Δd'<Δd, to obtain a new normal ratio, and the above steps are repeated until the normal ratio is greater than or equal to 0.3. The value on the y-axis corresponding to the upper straight line of the two straight lines is used as the target concentration.

[0010] As a further solution of the present invention: when the value on the y-axis corresponding to the upper straight line of the two straight lines is less than the preset value, if the target concentration is still not determined, the judgment standard is lowered from 0.3 to 0.25, and the target concentration is re-determined.

[0011] As a further solution of the present invention: after obtaining the predicted concentration, the method further includes: The actual smoke concentration generated by the metal currently being smelted is obtained and recorded as the actual concentration. When the difference between the predicted concentration and the actual concentration is greater than a preset difference threshold, a prompt message is sent to report an error.

[0012] As a further solution of the present invention: the alarm information includes an audible alarm and an optical alarm.

[0013] The beneficial effects of the present invention are as follows: The present invention uses a machine learning model to screen elements that have a significant impact on smoke concentration, only monitors necessary target elements, and groups them for melting at differentiated content intervals to form a uniform calibration sample, and obtains the furnace smoke curve in real time by laser scattering; then adopts the "maximum value-translation straight line" method, and takes the peak value as the target concentration only when the concentration is continuously close to the peak value for a long period of time, so as to avoid the high target concentration caused by instantaneous spikes; after mapping and fitting the target element content of all groups with the target concentration, the potential smoke level can be predicted before melting and the pollution risk can be prompted; during the melting process, the measured concentration is compared with the predicted concentration. If the deviation is abnormal, an error is reported. At the same time, when the predicted or measured concentration exceeds the threshold, an audible and visual alarm is triggered synchronously, thereby achieving advance warning and real-time protection against medium frequency furnace smoke pollution, and improving the reliability of air quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 The present invention is a flow chart of an intelligent monitoring and alarm system for a medium frequency furnace. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] See also Figure 1 As shown, the present invention is an intelligent monitoring and alarm system for a medium frequency furnace, comprising: Setting module: select the target element and set the content range of the target element; In a preferred embodiment of the present invention, selecting the target element includes: Marking elements contained in the metal as initial elements, determining the degree of influence of the initial elements on smoke concentration based on a random forest model and / or a gradient boosting tree, and selecting the initial elements whose influence is greater than a preset influence threshold as target elements; For example, a historical smelting condition database is first established, and the chemical composition test results of each heat of metal are associated with the corresponding laser scattering smoke concentration curve to generate a sample set containing element content vectors and peak concentration labels; then, the missing items are interpolated by means and the content data of different dimensions are standardized. With the content of each element as the feature and the peak smoke concentration as the target value, a random forest model and a gradient boosting tree model are trained respectively using grid search cross-validation. After the model training is completed, the feature importance score of each element in the two types of models is extracted, and the importance of the two types of models is normalized and the average value is taken as the comprehensive influence; then, the comprehensive influence is sorted from high to low, and the process experts give the influence threshold based on the common element sources and post-smelting quality requirements. For example, if the comprehensive influence of chromium, manganese and silicon in the iron-based alloy is higher than the threshold, chromium, manganese and silicon are calibrated as target elements; for the content range of each target element, the historical data distribution can be referred to, combined with national standards, furnace lining corrosion resistance and product performance requirements, and a certain process safety margin can be left between the minimum and maximum values; It is understandable that by establishing a correlation between historical smelting data and the smoke concentration curve and introducing random forest and gradient boosting tree models for feature importance analysis, the true impact of each element on smoke generation can be revealed based on objective data, avoiding omissions or redundancies caused by selecting factors based solely on experience. On this basis, process personnel set the impact threshold and content range based on production requirements, so that subsequent calibration experiments are concentrated on the most representative elements and within a reasonable content range. This not only reduces the test scale and monitoring load, but also ensures sample coverage and uniformity, thereby improving the adaptability of the fitting curve to different ratios and the credibility of the prediction results. Ultimately, the system can predict the risk of smoke exceeding the standard earlier and more accurately, laying a reliable foundation for timely ventilation or process adjustment measures on site. Calibration module: Set several content nodes within the content range, group the content nodes, select the target metal corresponding to group i, add the target metal into the intermediate frequency furnace for melting, and obtain the smoke concentration emitted by the intermediate frequency furnace in real time. The maximum smoke concentration within the preset monitoring time is used as the target concentration; In another preferred embodiment of the present invention, selecting the target metal includes: Starting from the starting point of the content range, a number of content nodes are set at preset content intervals, and different target elements correspond to different content intervals; Grouping the content nodes of different target elements, wherein different groups contain at least one different content node, and there is only one content node for one target element in the same group; Select a target metal, where the content of the target element in the target metal is the content corresponding to the content node in group i; In specific implementation, first check the starting point and end point of the established content range of the target element, and set a unique content interval for each factor based on its sensitivity to smoke and process controllable accuracy, such as a step of 0.2% for chromium and a step of 0.5% for manganese, and then accumulate the intervals from the starting point to obtain several content nodes; then use the Latin square idea to combine the nodes of each factor into groups, ensuring that each group only takes one node for a certain factor and there is at least one different content between groups, such as the first group takes the lowest node for chromium, the middle node for manganese, and the highest node for silicon, and the second group only increases chromium by one node while keeping other factors unchanged. The groups constructed in this way cover the entire interval and avoid combination redundancy; then weigh the raw materials according to the node content of each group and mix them into the target metal, put them into the medium frequency furnace and smelt them according to the standard heating curve, and use a laser scattering instrument to continuously record the smoke concentration at a frequency of seconds before melting to steel tapping. After the preset monitoring period, the highest point is extracted from the complete curve as the target concentration corresponding to the heat, providing a one-to-one sample for subsequent multi-dimensional coordinate fitting; Setting differentiated nodes and conducting grouped experiments in this way, on the one hand, forms representative samples within the raw material ratio space through a uniform and limited parameter grid, avoiding the time and cost burden of traversing all possible combinations. On the other hand, it ensures that only a single or a few factors are fine-tuned during each smelting process. This helps to observe the direct impact of individual elements on smoke peaks without changing other conditions, minimizing the interference of multi-factor coupling. The highest concentration measured in real time is used as the target value, which is equivalent to capturing the most demanding operating conditions, making the fitting model robust to fluctuations in actual production. Therefore, the entire process lays the data foundation for establishing a reliable mapping between content and smoke concentration, ultimately achieving early identification of potential exceedance risks within the normal production ratio range and triggering timely warnings. In another preferred embodiment of the present invention, obtaining the smoke concentration includes: Installing a laser scattering monitoring device at a preset position to obtain the smoke concentration based on the reflection of the laser; It is worth noting that, firstly, a transmitting end and a receiving end are arranged in the furnace exhaust channel to form a corresponding optical path. The transmitting end uses a near-infrared semiconductor laser of a selected wavelength to continuously output a constant power beam. After entering the flue gas through the quartz window, the suspended particles produce Mie scattering and Rayleigh scattering on the laser. The scattered light is received by the silicon photodiode array at a monitoring angle perpendicular to the incident light. The receiving end first converts the detection signal into current-voltage, and then uses a phase-locked amplifier to filter out the low-frequency noise generated by the fire pulsation to obtain a digital voltage value proportional to the instantaneous scattered light intensity. The system collects the data from the signal buffer at fixed sampling periods. The algorithm extracts a frame of raw values from the burst zone, first performing a zero-point calibration using the empty furnace gas as a benchmark, then calling a preset particle concentration-scattering intensity calibration curve to back-calculate the light intensity after temperature drift compensation into the smoke concentration. To reduce the impact of particle size distribution fluctuations, the algorithm performs median filtering on the concentration results of several consecutive frames within a moving time window and outputs a smoothed value, while recording the maximum value for target concentration extraction. For example, in a single smelting operation, if the laser scattering echo shows a concentration increase over multiple consecutive frames after model conversion, the system will write the highest point into the database in real time and push it to the host computer, completing the dynamic acquisition of the smoke concentration of that furnace. In another preferred embodiment of the present invention, the process of obtaining the target concentration further includes: Draw a curve A showing the change of smoke concentration over time during the monitoring period; Draw a straight line through (0, max(f(t))) and parallel to the x-axis, and draw a straight line through (0, max(f(t))-Δd) and parallel to the x-axis, where Δd is a first preset value; The portion of the curve A between the two straight lines is recorded as the normal portion, and the proportion of the domain of the normal portion to the monitoring time is calculated and recorded as the normal proportion; If the normal ratio is less than 0.3, both straight lines are shifted downward by Δd', where Δd' is a second preset value and Δd'<Δd, to obtain a new normal ratio, and the above steps are repeated until the normal ratio is greater than or equal to 0.3. The value on the y-axis corresponding to the upper straight line of the two straight lines is used as the target concentration; In a preferred embodiment of the present invention, when the value on the y-axis corresponding to the upper straight line of the two straight lines is less than the preset value, if the target concentration has not yet been determined, the judgment standard is lowered from 0.3 to 0.25, and the target concentration is re-determined; It should be noted that the use of two straight lines parallel to the time axis to perform window screening on the smoke curve is to eliminate the peak value that only appears at an instant and extract the steady-state interval that can represent the high emission state of the furnace: the first straight line is located at the vertical coordinate corresponding to the smoke peak, and the second straight line maintains a fixed vertical distance Δd from it (such as 10mg / m 3), when the time between the two lines is less than 30% of the monitoring time, it means that the peak may be caused by occasional disturbances, so the two lines are synchronously shifted down by a smaller step size Δd' (such as 3mg / m 3 ) and recalculate the ratio, repeating the iteration until the duration of the high interval of the curve after the reduction reaches or exceeds 30%, and then setting the vertical coordinate of the upper straight line as the target concentration; if the upper straight line is still lower than the preset minimum reference concentration after multiple downward shifts and the target value cannot be determined, the duration criterion is appropriately relaxed to one-quarter to ensure that a representative concentration point can be selected. This not only avoids mistaking accidental peaks for target concentrations, resulting in subsequent high fitting, but also prevents the loss of low-emission samples due to over-tightening conditions, thereby providing consistent, robust and comprehensive data support for multi-dimensional fitting, helping the system to accurately predict the risk of smoke exceeding the standard under various ratios; Generate coordinate points (C1, C2, ..., C n , D), C n represents the content of the target element corresponding to the nth dimension, D represents the target concentration, and n represents the total number of the target elements; Generate coordinate points corresponding to all groups, and perform fitting on all the coordinate points to obtain a fitting curve; During the specific implementation process, the element content and corresponding target concentration of each group of target metals are first read from the calibration experiment database in sequence, and the content of each element is mapped to the first axis, the second axis, the third axis, and so on in order, and the target concentration is mapped to the last vertical axis, thereby forming a set of coordinate points with clear physical meaning in the multidimensional space; then the same mapping process is repeated for the next group of proportions until all groups are converted into coordinate points, and the system generates a sample set containing all points in the memory; then the library function is called to select the support vector regression method based on the radial basis kernel, and the multidimensional content is used as the independent variable and the target concentration as the dependent variable, and the fitting is completed on the sample set. The algorithm automatically adjusts the penalty coefficient and the kernel function width through a five-fold cross-validation to avoid overfitting; after the fitting is completed, a mapping model between the element content and the predicted concentration is obtained, and the system serializes and stores the model on the server for subsequent calls; By employing multidimensional coordinate mapping and regression fitting, the relationship between the mutually coupled element contents and smoke concentration can be integrated into the same mathematical framework. This allows the prediction phase to instantly obtain the corresponding concentration estimate simply by inputting the actual mix ratio, thus providing a fast and unified calculation entry point for pre-smelting risk assessment. Furthermore, the nonlinear mapping introduced by the kernel function can capture the potential interaction effects between elements, ensuring that the model remains reliably applicable to different mix ratio combinations. Ultimately, this lays the data and algorithmic foundation for the system to issue timely warnings of smoke exceeding the standard. Alarm module: obtains the content of the target element in the metal currently being smelted, records it as the comparative content, substitutes the comparative content into the fitting curve to obtain the predicted concentration, and sends an alarm message when the predicted concentration is greater than a preset concentration threshold; It should be noted that the alarm information includes sound alarm and optical alarm; It is understandable that before the charge is added and the temperature is raised, the spectrometer is called to sample the molten iron, and the spark direct reading spectroscopy is used to obtain the real-time content results of each target element within tens of seconds. The results are loaded into the host computer database and immediately extracted by the background script, and the missing items are corrected by nearest neighbor interpolation and composed into a vector in the input order required by the fitting model. The vector is then input into the previously serialized support vector regression model to obtain the predicted concentration; the monitoring thread then compares the predicted concentration with the concentration threshold set by the process department. When the predicted value exceeds the threshold, the alarm routine is triggered, and a Bohr signal is sent to the on-site PLC through the industrial Ethernet. At the same time, a red dialog box pops up on the operating table and the buzzer module and the high-brightness warning light are called to realize the double sound and light reminder. If the predicted value does not exceed the limit, the next component detection data is periodically waited for to repeat the above process; Substituting real-time composition into the fitted model to estimate dust concentration in advance at the initial stage of smelting allows operators to be aware of potential risk of exceeding standards before concentration actually rises, allowing time to adjust ingredients or activate enhanced ventilation, rather than passively taking action after high concentrations are measured. By triggering audible and visual alarms by comparing with fixed thresholds, this provides intuitive and uninterrupted prompts in noisy smelting sites, reducing missed calls due to human negligence. Overall, this creates a closed loop between the system's predictive capabilities and on-site response, laying the foundation for proactive protection against dust pollution and process safety management. Another preferred embodiment of the present invention further comprises: Obtaining the actual smoke concentration generated by the metal currently being smelted, recording it as the actual concentration; when the difference between the predicted concentration and the actual concentration is greater than a preset difference threshold, sending a prompt message to report an error; It is worth noting that continuing to monitor and calculate in real time the gap between the measured concentration and the predicted value after the predicted concentration is generated can add a self-checking mechanism to the system. When the deviation between the two exceeds the preset threshold, an error will be immediately reported. This can not only timely detect hardware anomalies such as sensor failure and sampling tube blockage, but also expose model inaccuracies caused by sudden ingredient deviations or sudden changes in furnace conditions, thereby prompting on-site personnel to check the equipment status or re-measure the composition and update the model coefficients. This makes the entire monitoring process reliable in a dynamic production environment, avoids the impact of false negative omissions or false positive alarms on operational decisions, and further consolidates the system's ability to predict and warn of the risk of excessive smoke.

[0018] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An intelligent monitoring and alarm system for a medium frequency furnace, characterized in that: include: Setting module: select the target element and set the content range of the target element; Calibration module: Set several content nodes within the content range, group the content nodes, select the target metal corresponding to group i, add the target metal into the intermediate frequency furnace for melting, and obtain the smoke concentration emitted by the intermediate frequency furnace in real time. The maximum smoke concentration within the preset monitoring time is used as the target concentration; Generate coordinate points (C1, C2, ..., C n , D), C n represents the content of the target element corresponding to the nth dimension, D represents the target concentration, and n represents the total number of the target elements; Generate coordinate points corresponding to all groups, and perform fitting on all the coordinate points to obtain a fitting curve; Alarm module: obtains the content of the target element in the metal currently being smelted, records it as the comparative content, substitutes the comparative content into the fitting curve to obtain the predicted concentration, and sends an alarm message when the predicted concentration is greater than a preset concentration threshold.

2. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: Select target elements include: Elements contained in the metal are marked as initial elements, the degree of influence of the initial elements on smoke concentration is determined based on a random forest model and / or a gradient boosting tree, and the initial elements whose influence is greater than a preset influence threshold are used as target elements.

3. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: Selected target metals include: Starting from the starting point of the content range, a number of content nodes are set at preset content intervals, and different target elements correspond to different content intervals; Grouping the content nodes of different target elements, wherein different groups contain at least one different content node, and there is only one content node for one target element in the same group; A target metal is selected, wherein the content of the target element in the target metal is the content corresponding to the content node in group i.

4. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: Obtaining smoke concentration includes: A laser scattering monitoring device is installed at a preset position to obtain the smoke concentration based on the reflection of the laser.

5. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: The process of obtaining the target concentration further includes: Draw a curve A showing the change of smoke concentration over time during the monitoring period; Draw a straight line through (0, max(f(t))) and parallel to the x-axis, and draw a straight line through (0, max(f(t))-Δd) and parallel to the x-axis, where Δd is a first preset value; The portion of the curve A between the two straight lines is recorded as the normal portion, and the proportion of the domain of the normal portion to the monitoring time is calculated and recorded as the normal proportion; If the normal ratio is less than 0.3, both straight lines are shifted downward by Δd', where Δd' is a second preset value and Δd'<Δd, to obtain a new normal ratio, and the above steps are repeated until the normal ratio is greater than or equal to 0.

3. The value on the y-axis corresponding to the upper straight line of the two straight lines is used as the target concentration.

6. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 5, characterized in that: When the value on the y-axis corresponding to the upper straight line of the two straight lines is less than the preset value, if the target concentration has not yet been determined, the judgment standard is lowered from 0.3 to 0.25, and the target concentration is re-determined.

7. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: After obtaining the predicted concentration, it also includes: The actual smoke concentration generated by the metal currently being smelted is obtained and recorded as the actual concentration. When the difference between the predicted concentration and the actual concentration is greater than a preset difference threshold, a prompt message is sent to report an error.

8. The intelligent monitoring and alarm system for a medium frequency furnace according to claim 1, characterized in that: The alarm information includes an audible alarm and an optical alarm.

Images