Claims:1. A learning system for manufacturing process optimization is provided with restricted Boltzmann machines and a multi-objective evolutionary model to yield optimized process factors comprising:
a. Material selection based on requirement and execution of the experiment on the WEDM machine;
b. A novel hybrid learning model to generate the optimized process factors;
c. Calculate and execute the layers in accordance with the desired accuracy.
2. The system as claimed in claim 1, wherein the hybrid model's first component is a restricted Boltzmann machine and the second component is a Multi Objective evolutionary model that blends two different models.
3. The system as claimed in claim 1, wherein the blends deep learning and unsupervised learning.
4. The system as claimed in claim 1, wherein the system is hybrid and further needs to be executed in sequence to provide the output. The components are not configured to change the sequence of the execution.
5. The system as claimed in claim 1, wherein the generate multi objective optimization values at a same time. The proposed model generated material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) optimal factors.
6. The system as claimed in claim 1, wherein the system is capable of generating optimal factors data for a relatively small data collection.
7. The system as claimed in claim 1, wherein the system is outperformed Response Surface Methodology (RSM) on the low surface response factors.
8. The system as claimed in claim 1, wherein the system is outperformed other conventional machine learning models - ANFIS, ANN and SVM
9. The system as claimed in claim 1, wherein the system is achieved 98.8 percent, 97.6 percent and 98.2 percent of accuracy for generating optimized manufacturing factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) respectively.
10. The system as claimed in claim 1, wherein the system enables the reduction of production time and set-up time, as well as the reduction of costs associated with WEDM processing while increasing productivity, and it also enables automation, and further, implemented to calculate process parameters for other non-standard machining processes as well as for other metals.
, Description:[001] The present invention provides process, methods, and systems for optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model.
Background of the invention
[002] Modern manufacturing processes are extremely complicated, owing to the presence of a large number of interrelated process variables. Recent manufacturing processes are complex, restricted, non-convex, multi-response, and dependent on several process variables, making optimization challenging. In manufacturing, developing optimum, correlational, or functional models is a demanding task.
[003] Wire electrical discharge machining (WEDM) has emerged as a significant non-traditional machining technology in recent years. WEDM can play a significant role in some production industries due to its ability to cut complicated and intricate shapes of components precisely and accurately in all electrically conductive materials. Some of the challenges of WEDM is it requires high operational costs, lengthy experimentation periods, and the need for qualified machinists. Reduced operational expenses, experimental periods, and the elimination of the requirement for qualified machinists all contribute to WEDM's increased reach, which benefits the manufacturing industry significantly.
[004] To enhance the manufacturing capability, process parameters must be optimized. However, optimizing process parameters with the fewest possible experiments is always a challenge. With the growth of Artificial intelligence, machine learning and deep learning optimization of process parameters with minimal experiments is always a possibility. Certain methods, such as Taguchi and Response Surface methodology, were already available, but obtaining the optimal process parameters with the fewest possible experiments has always been challenging even with these methods.
[005] Machine learning models come in three flavors: supervised learning, unsupervised learning, and reinforcement learning. Supervised learning provides the learner with labelled training instances that have the correct input and output. By assigning no labels to data points, unsupervised learning recognizes and explores relationships in data. In reinforcement learning, maps conditions to behavior to optimize a scalar reward. The proposed learning model is a hybrid model which combines deep learning with unsupervised learning.
[006] The present model achieved 98.8 percent, 97.6 percent, and 98.2 percent of accuracy for generating optimized manufacturing factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) respectively.
[007] The conventional models, while establishing hybrid complex models is always challenging, this indicate that hybrid algorithms are more efficient because they can exchange characteristics to mitigate their weaknesses and maximize their advantages. The novel learning model proposed here is a hybrid model composed of two components. The first component is a probability distribution dataset generated using a modified RBM, and the second component is a multi-objective evolutionary model used to generate outcomes.
[008] Hence, these above-mentioned technical problems are to be solved by the present invention by providing a unique model which is about optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model. The present disclosure describes an intelligent method for modelling and optimizing the wire electrical discharge machining (WEDM) process.
[009] It is an object of the present invention to provide an improved process, method, and system to address the existing challenge. Alternatively, it is an object of the present invention to address the foregoing problems or at least to provide the public with a useful choice.
Summary of the present invention
[010] The present invention provides process, methods, and systems for optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model.
[011] The present disclosure is data-driven and is applicable to any manufacturing process and metal. The present invention is a novel learning model for optimizing manufacturing processes using deep learning models. This concept contributes to the elimination of high operational costs, lengthy experimentation periods, and the need for qualified machinists.
[012] In an embodiment, a novel learning model for manufacturing process optimization is presented that consists of modified restricted Boltzmann machines and a multi-objective evolutionary model to yield optimized process factors comprises a material selection based on requirement and execution of the experiment on the WEDM machine, a novel hybrid learning model to generate the optimized process factors and calculate and execute the layers in accordance with the desired accuracy.
[013] The model of the present invention, where the hybrid model's first component is a restricted Boltzmann machine, and the second component is a Multi Objective evolutionary model that blends two different models. The present model blends deep learning and unsupervised learning. The hybrid learning model, needs to be executed in sequence to provide the output. The components are not configured to change the sequence of the execution. The model can generate multi objective optimization values at a same time. The proposed model generated material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) optimal factors. Further, the model described can generate optimal factors data for a relatively small data collection. The present model outperformed Response Surface Methodology (RSM) on the low surface response factors. The present model outperformed other conventional machine learning models - ANFIS, ANN and SVM.
[014] The proposed model achieved 98.8 percent, 97.6 percent and 98.2 percent of accuracy for generating optimized manufacturing factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) respectively. The proposed model enables the reduction of production time and set-up time, as well as the reduction of costs associated with WEDM processing while increasing productivity, and it also enables automation. This model can be developed and implemented to calculate process parameters for other non-standard machining processes as well as for other metals.
[015] The innovative description is a novel learning model comprising of start, preliminary design of the metal, WEDM experimentation design, WEDM experimentation implementation, WEDM experimentation analysis, probability distribution dataset using RBM, generate results using multi objective evolutionary model, create RSM, compare results between RSM and multi objective evolutionary model, result analysis and stop.
[016] The present model uses modified Restricted Boltzmann Machine (RBM) and a multi objective evolutionary model to optimize the machining factors. The processing model as described in drawing of the present invention provides an example of the system in which the system components and techniques listed can be implemented on a machining product (WEDM) and computer programs that can run on one or more WEDM/computers and any material.
[017] The preliminary design of the metal begins with the material selection. A metal is selected based on the requirements of the application, such as dental implants or computer chips. Metal choices must be generic in nature and must satisfy all fundamental needs such as cost, durability, and so forth. Metal selection factors are also depending on material combinations. Material combinations include resin composites, cements, glass ionomers, ceramics, noble and base metals, amalgam alloys, gypsum materials, gold, various titanium grades, and zirconium.
[018] The design of a WEDM experiment entails selecting input parameters such as pulse on time, pulse off time, peak current, wire feed rate, wire tension, and servo feed. The output of the process factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) are dependent on the input parameters. To obtain the best results, the input parameters must be properly combined.
[019] WEDM experimentation is carried out by executing the selected material on the machine. The execution is determined by the input parameters, which include the pulse on time, the pulse off time, the peak current, the wire feed rate, the wire tension, and the servo feed. While experiments can be performed endlessly in principle, this invention requires that the experiment be done only 22 times. The experiment's input parameters are adjusted on a per-run basis.
[020] WEDM analysis entails evaluating the following process variables: material removal rate (RMM), surface roughness (SR), and wire wear ratio (WWR). Further analysis of the discovered process parameters is required to establish whether the optimal state occurred. The obtained data is passed on to the next component Response Surface Methodology (RSM) to calculate the surface roughness.
[021] Further, the present art is about optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model. The present invention describes an intelligent method for modelling and optimizing the wire electrical discharge machining (WEDM) process. Titanium alloys are preferred in a variety of industries, including dentistry and computer chips, due to their superior strength, durability, corrosion resistance, strength to density ratio, and light weight. Precision manufacturing, particularly in the dental industry, is critical. Precision machining and manufacturing using traditional methods is extremely time consuming, and alternative methods such as WEDM can be used. Calculating the material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) experimentally is time consuming and costly. These costs and time losses can be avoided by accurately forecasting the output quality characteristics of the selected input machining factors. The present disclosure includes a model for multi-object optimization that is divided into two subsystems. The first subsystem, Restricted Boltzmann Machine, is constructed and trained using dominant solutions in order to implicitly extract the distributed representative information of the decision variables in the viable subset. The RBM sampling algorithm is updated on a periodic basis to provide the probability of sampling subsets. After obtaining the optimal manufacturing data, the second subsystem's multi objective evolutionary model will be invoked. The final manufacturing process optimization factors will be determined by the multi objective evolutionary model. The obtained solutions are validated using boundary conditions, and statistical analysis demonstrates the presented model's reliability and accuracy.
[022] Other objects and advantages of the present invention will become apparent from the following detailed description when viewed in conjunction with the accompanying drawings, which set forth certain embodiments of the invention.
Brief description of the Drawings
[023] FIG. 1-2 is a flowchart depicting the overall process flow of the present invention.
Detailed Description of the present Invention
[024] Description of the Invention for a thorough understanding of the present invention, reference is made to the following detailed description in connection with the abovementioned drawings. Although the present invention is described with reference to exemplary embodiments, the present invention is not intended to be limited to the specific forms set forth herein. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but these are intended to cover the application or implementation without departing from the spirit or scope of the present invention.
[025] Further, it will nevertheless be understood that no limitation in the scope of the invention is thereby intended, such alterations and further modifications in the figures and such further applications of the principles of the invention as illustrated herein being contemplated as would normally occur to one skilled in the art to which the invention relates.
[026] Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. Further, reference herein to “one embodiment” or “an embodiment” means that a particular feature, characteristic, or function described in connection with the embodiment is included in at least one embodiment of the invention.
[027] Furthermore, the appearances of such phrase at various places herein are not necessarily all referring to the same embodiment. The terms “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced items.
[028] The present invention by providing a unique model which is about optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model. The present disclosure describes an intelligent method for modelling and optimizing the wire electrical discharge machining (WEDM) process.
[029] The present disclosure is data-driven and is applicable to any manufacturing process and metal. The present invention is a novel learning model for optimizing manufacturing processes using deep learning models. This concept contributes to the elimination of high operational costs, lengthy experimentation periods, and the need for qualified machinists.
[030] In an embodiment, a novel learning model for manufacturing process optimization is presented that consists of modified restricted Boltzmann machines and a multi-objective evolutionary model to yield optimized process factors comprises a material selection based on requirement and execution of the experiment on the WEDM machine, a novel hybrid learning model to generate the optimized process factors and calculate and execute the layers in accordance with the desired accuracy.
[031] The model of the present invention, where the hybrid model's first component is a restricted Boltzmann machine, and the second component is a Multi Objective evolutionary model that blends two different models. The present model blends deep learning and unsupervised learning. The hybrid learning model, needs to be executed in sequence to provide the output. The components are not configured to change the sequence of the execution. The model can generate multi objective optimization values at a same time. The proposed model generated material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) optimal factors. Further, the model described can generate optimal factors data for a relatively small data collection. The present model outperformed Response Surface Methodology (RSM) on the low surface response factors. The present model outperformed other conventional machine learning models - ANFIS, ANN and SVM.
[032] The proposed model achieved 98.8 percent, 97.6 percent and 98.2 percent of accuracy for generating optimized manufacturing factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) respectively. The proposed model enables the reduction of production time and set-up time, as well as the reduction of costs associated with WEDM processing while increasing productivity, and it also enables automation. This model can be developed and implemented to calculate process parameters for other non-standard machining processes as well as for other metals.
[033] Further, the present art is about optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model. The present invention describes an intelligent method for modelling and optimizing the wire electrical discharge machining (WEDM) process. Titanium alloys are preferred in a variety of industries, including dentistry and computer chips, due to their superior strength, durability, corrosion resistance, strength to density ratio, and light weight. Precision manufacturing, particularly in the dental industry, is critical. Precision machining and manufacturing using traditional methods is extremely time consuming, and alternative methods such as WEDM can be used. Calculating the material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) experimentally is time consuming and costly. These costs and time losses can be avoided by accurately forecasting the output quality characteristics of the selected input machining factors. The present disclosure includes a model for multi-object optimization that is divided into two subsystems. The first subsystem, Restricted Boltzmann Machine, is constructed and trained using dominant solutions in order to implicitly extract the distributed representative information of the decision variables in the viable subset. The RBM sampling algorithm is updated on a periodic basis to provide the probability of sampling subsets. After obtaining the optimal manufacturing data, the second subsystem's multi objective evolutionary model will be invoked. The final manufacturing process optimization factors will be determined by the multi objective evolutionary model. The obtained solutions are validated using boundary conditions, and statistical analysis demonstrates the presented model's reliability and accuracy.
[034] In an embodiment, the innovative description is a novel learning model comprising of start, preliminary design of the metal, WEDM experimentation design, WEDM experimentation implementation, WEDM experimentation analysis, probability distribution dataset using RBM, generate results using multi objective evolutionary model, create RSM, compare results between RSM and multi objective evolutionary model, result analysis and stop. This is also shown in the figure.1 of the present invention, a step-by-step procedure of the hybrid model.
[035] The present model uses modified Restricted Boltzmann Machine (RBM) and a multi objective evolutionary model to optimize the machining factors. The processing model as described in drawing of the present invention provides an example of the system in which the system components and techniques listed can be implemented on a machining product (WEDM) and computer programs that can run on one or more WEDM/computers and any material.
[036] The preliminary design of the metal begins with the material selection. A metal is selected based on the requirements of the application, such as dental implants or computer chips. Metal choices must be generic in nature and must satisfy all fundamental needs such as cost, durability, and so forth. Metal selection factors are also depending on material combinations. Material combinations include resin composites, cements, glass ionomers, ceramics, noble and base metals, amalgam alloys, gypsum materials, gold, various titanium grades, and zirconium.
[037] The design of a WEDM experiment entails selecting input parameters such as pulse on time, pulse off time, peak current, wire feed rate, wire tension, and servo feed. The output of the process factors - material removal rate (RMM), the surface roughness (SR), and the wire wear ratio (WWR) are dependent on the input parameters. To obtain the best results, the input parameters must be properly combined.
[038] WEDM experimentation is carried out by executing the selected material on the machine. The execution is determined by the input parameters, which include the pulse on time, the pulse off time, the peak current, the wire feed rate, the wire tension, and the servo feed. While experiments can be performed endlessly in principle, this invention requires that the experiment be done only 22 times. The experiment's input parameters are adjusted on a per-run basis.
[039] WEDM analysis entails evaluating the following process variables: material removal rate (RMM), surface roughness (SR), and wire wear ratio (WWR). Further analysis of the discovered process parameters is required to establish whether the optimal state occurred. The obtained data is passed on to the next component Response Surface Methodology (RSM) to calculate the surface roughness.
[040] In another embodiment, the conventional models, while establishing hybrid complex models is always challenging, this indicate that hybrid algorithms are more efficient because they can exchange characteristics to mitigate their weaknesses and maximize their advantages. The novel learning model proposed here is a hybrid model composed of two components. The first component is a probability distribution dataset generated using a modified RBM, and the second component is a multi-objective evolutionary model used to generate outcomes.
[041] The proposed modified RBM works in three phases. The first phase forward pass, the second phase being backward pass and the final phase being reconstruction of the required dataset. In the forward pass state, all visible nodes' inputs are transferred to all hidden nodes. Following that, the non-linear activation function is invoked. The hyperbolic tangent Activation Function is used in the present disclosure. The value of each input node is multiplied by the supplied weights and added to the hidden layer. The result is feed forwarded to the hyperbolic tangent activation function. The function provides is output of the first phase. The backward pass step distributes the results from the first phase using a random probability distribution value set. After validating the samples, the negative gradient values are discarded. The visible layer is then redirected to the visible layer using the positive values. Iterate through phases one and two until ideal values are obtained. In the reconstruction phase, the model assigns a probability to each possible binary state vectors over its visible units. The weight matrix, values from phase two are go through the sigmoid function. The final output is then produced.
[042] After the modified restricted Boltzmann machine generates values, the values are passed to the multi objective evolutionary model, which then begins executing. The modified multi objective evolutionary model is implemented as follows:
ALGORITHM 1.

1. Initial Generation
2. Initial parameters
3. Generate reference points
4. Generate initial population
5. Generate Ideal points
6. Analyze the initial population using fitness functions
7. Detect the non-dominated population and sort accordingly
8. While all iterations:
1. Randomize the population
2. Using tournament methods, choose two parents P1, P2.
3. Apply crossover between the chosen parents
4. Apply the mutation on the chosen parents
5. Identify the non-dominated population
6. Sort the non-dominated population
7. Normalize the population members with the reference points identified in step e
8. Apply the niche preservation
9. Acquire and transmit on to the next generation members of the niche-acquired population
9. Publish the predication values

[043] The present model is comprised of eight primary steps, each of which is crucial to the model's implementation. The proposed model begins with the initial generation, which would initialize the total population obtained from the prior component – restricted Boltzmann machine. The second step is to initialize the model's parameters. In the fourth phase, a random function is used to generate an initial population of size N based on the predefined set of reference points selected in the third step.
[044] A two-layer technique is used to generate the reference points. The first layer of reference points generate boundary points, whereas the second layer generates inner boundary points. Then, the reference points for the inner boundary are transferred to the boundary points, and the final reference points are established. Based on the initial population, the ideal points are constructed in step five.
[045] The non-dominated population is identified and classified using steps six and seven. Step seven applies the tournament selection technique and the default fitness methods/functions to the initially selected population. In step seven, the non-dominant population is detected and classified appropriately. Iterate through Step 8 until the stopping criteria is met. The terminating criteria could be the number of iterations required by the algorithm or the number of echoes required by the overall method. A random population is found and selected in step a. In step b, the tournament approach is used to pick two individuals, p1 and p2. If the selected individuals share same front, then new individuals are chosen. This phase would be repeated until individuals from different fronts are chosen.
[046] Crossover and mutation are applied to the selected p1, p2 individuals in steps c, d. Individuals are assigned mutation probabilities during the crossover step. Thereafter, a random cut point is applied to each parent's subdivisions. The subdivisions are then exchanged, and the correction would be applied. In steps e, f the population is updated based on the generic operators applying and sorting criteria. The sorting criteria is fronts and crowding distance. The Normalized step g creates an initial vector for the population by selecting the minimal value for each objective function. Then, each objective function is translated to calculate the needed vector's extreme points.
[047] The niche preservation step h identifies the reference point with the shortest argument life. If there are multiple reference points, a random reference point is chosen. The remaining reference points are deleted, and the members linked with the reference point with the smallest perpendicular distance are chosen. The niche count is then incremented and preserved based on the reference point.
[048] The final stage i is to preserve the niche population and pass it on to the next generation of members. Members of the subsequent generation would use the niche population values to generate their own values. Finally, the generated values are published. The results from the hybrid model are validated by comparing them to those values generated by the RSM model and the final values would be available.
[049] It is accordingly an object of the present invention to provide a novel deep learning algorithm was evaluated using the Titanium metal alloy WEDM dataset. Two phases of evaluation are used to assess the model. In the first phase, the model is compared to a widely utilized manufacturing process, namely the response surface methodology (RSM).
[050] Hence, the present invention is providing a unique model which is about optimizing manufacturing processes using a hybrid deep learning model composed of a Restricted Boltzmann Machine (RBM) and a multi-objective evolutionary model. The present disclosure describes an intelligent method for modelling and optimizing the wire electrical discharge machining (WEDM) process.
[051] What has been described above includes examples of the subject invention. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the subject invention, but one of ordinary skill in the art may recognize that many further combinations and permutations of the subject invention are possible. Accordingly, the subject invention is intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims. Furthermore, to the extent that the term “includes” is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term “comprising” as “comprising” is interpreted when employed as a transitional word in a claim.
[052] The foregoing descriptions of specific exemplary embodiments of the present invention have been presented for purposes of illustration and description only. They are not intended to be exhaustive or to limit the invention to the precise forms and sequence of steps disclosed, and obviously many modifications and variations are possible considering the above teachings.
[053] The exemplary embodiments were chosen and described in order to explain certain principles of the invention and their practical application, thereby enabling others skilled in the art to make and utilize various exemplary embodiments of the present invention, as well as various alternatives and modifications thereof.